17.6 – ANCOVA – analysis of covariance

Introduction

Analysis of covariance (ANCOVA) is intended to help with analysis of designs with categorical treatment variables on some response (dependent) variable, but a known confounding variable is also present. Thus, the researcher is also likely to know of additional ratio scale variables that covary with the response variable and, moreover, must be included in the experimental design in some way.

Take for example the well-known relationship between body size and whole-animal metabolic rate as measured by rates of carbon dioxide production or rates of oxygen consumption for aerobic organisms. We may be interested in how addition or blocking of stress hormones affects resting metabolism; we may be interested in comparing men and women for activity metabolism, and so on. We’d like to know if the regressions were the same (eg, metabolic rate covaried with body mass in the same way — that is, the slope of the relationship was the same).

This situation arises frequently in biology. For example, we might want to know if male and female birds have different mean field metabolic rates, in which case we might be tempted to use a one-way ANOVA or t-test (since one factor with two levels). However, if males and females also differ for body size, then any differences we might see in metabolic rate could be due to differences in metabolic rate are confounded by differences in the covariable body size. We already discussed one approach to correction: calculate a ratio. Thus, a logical approach to correcting or normalizing for the covariation would be to divide body mass (units of kilograms) into metabolic rate (eg, volume of oxygen, O₂, consumed), and make comparisons, say, among different species, on mass-specific trait (). However, because the regression between mass and metabolic rate is allometric, i.e., not equal to one, the ratio does not, in fact normalize for body mass.  We made this point in Chapter 6.2, and remarked that analysis of covariance ANCOVA was a solution.

ANCOVA allows you to test for mean differences in traits like metabolic rate between two or more groups, but only after first accounting for covariation due to another variable (eg, body size). However, ANCOVA makes the assumption that relationship between the covariable and the response variable is the same in the two groups. This is the same as saying that the regression slopes are the same. We discussed how to use t-test to test hypothesis of equal slopes between regression models in Chapter 17.5, but a more elegant way is to include this in your model.

Example

We return to our sample of 13 tadpoles (Rana pipiens), hatched in the laboratory. I’ve repeated the data set in this page, scroll or click here.

Our linear model was and a scatterplot of the data set is shown in Figure 1 (a repeat of Figure 3 from Chapter 17.5, but now points identified to developmental group).

Figure 1. Scatterplot of oxygen consumption by R. pipiens tadpoles vs body mass (g) by developmental group (Gosner stages I or II).

The project looked at whether metabolism as measured by oxygen consumption was consistent across two developmental stages. Metamorphosis in frogs and other amphibians represents profound reorganization of the organism as the tadpole moves from water to air. Thus, we would predict some cost as evidenced by change in metabolism associated with later stages of development. Figure 2 shows a box plot of tadpole oxygen consumption by Gosner (1960) developmental stage (Figure 2 is a repeat of Figure 4 Chapter 17.5).

Figure 2. Copy of Figure 4, Chapter 17.5; boxplot of oxygen consumption of R. pipiens tadpoles by Gosner developmental stages.

Looking at Figure 2 we see a trend consistent with our prediction; developmental stage may be associated with increased metabolism. However, older tadpoles also tend to be larger, and the plot in Figure 2 does not account for that. Thus, body mass is a confounding variable in this example. There are several options for analysis here (eg, ANCOVA), but one way to view this is to compare the slopes for the two developmental stages. While this test does not compare the means, it does ask a related question: is there evidence of change in rate of oxygen consumption relative to body size between the two developmental stages? The assumption that the slopes are equal is a necessary step for conducting the ANCOVA.

So, divide the data set (Table 1) into two groups by developmental stage (12 tadpoles could be assigned to one of two developmental stages; one was at a lower Gosner stage than the others and so is dropped from the subset.

Table 1. and of 12 R. pipiens tadpole larvae by Gosner developmental stage.

Gosner stage I

1.76	109.41
1.88	329.06
1.95	82.35
2.13	198
2.26	607.7

Gosner stage II

Body mass
2.28	362.71
2.35	556.6
2.62	612.93
2.77	514.02
2.97	961.01
3.14	892.41
3.79	976.97

The slopes and standard errors we obtained in Chapter 17.5 were

Make a plot (Figure 3).

Figure 3. Scatterplot with best-fit regression lines of by for Gosner State I (closed circle, solid line) and Gosner Stage II (open circle, dashed line) R. pipiens tadpoles.

R code for plot in Figure 3.

#Used Rcmdr scatterplot(), then modified code
scatterplot(VO2~Body.mass | Gosner, regLine=FALSE, smooth=FALSE, 
boxplots=FALSE, xlab="Body mass (g)", ylab="Oxygen consumption (ml/h)", 
main="", cex=1.4, cex.axis=1.5, cex.lab=1.5, pch=c(19,19), by.groups=TRUE,
col=c("blue","red"), grid=FALSE, 
legend=list(coords="bottomright"), data=Tadpoles)
#Get regression equations for groups, subset by Gosner
abline(lm(VO2~Body.mass, data=Stage01), lty=1, lwd=2, col="blue")
abline(lm(VO2~Body.mass, data=Stage02), lty=1, lwd=2, col="red")

Returning to the important question, are the two slopes statistically indistinguishable (Ho: b_I = b_II), where b_I is the slope for the Gosner Stage I subset and b_II is the slope for the Gosner Stage II subset? We look at the plot, and since the lines cross, we tend to see a difference. Of course, we need to consider that our perception of slope differences may simply be chance, especially because the sample size is small. Proceed to test.

R code

The ANCOVA is a new ANOVA model where the factor variables are adjusted or corrected for the effects of the continuous variable.

R code for ANCOVA example, crossed or interaction model.

tadpole.1 <- lm(VO2 ~ Body.mass*Gosner, data=example.Tadpole)
summary(tadpole.1)
Anova(tadpole.1, type="II")

Output

summary(tadpole.1)

Call:
lm(formula = VO2 ~ Body.mass * Gosner, data = example.Tadpole)

Residuals:
    Min      1Q   Median      3Q      Max 
-167.80 -117.93    13.81   94.66   214.65

Coefficients:
                       Estimate   Std. Error   t value    Pr(>|t|) 
(Intercept)             -1231.6       783.0     -1.573    0.1544 
Body.mass                 750.0       390.7      1.919    0.0912 .
Gosner[T.II]              790.4       859.2      0.920    0.3845 
Body.mass:Gosner[T.II]   -350.1       409.5     -0.855    0.4174 
---
Signif. codes: 0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1

Residual standard error: 155.8 on 8 degrees of freedom
(1 observation deleted due to missingness)
Multiple R-squared: 0.821, Adjusted R-squared: 0.7539 
F-statistic: 12.23 on 3 and 8 DF, p-value: 0.002336

The coefficients for the first factor (GII) and then the differences in the coefficient for the second factor. You can just add the second coefficient to the first so they’re on the same scale.

Anova Table (Type II tests)

Response: VO2
                  Sum Sq   Df    F value     Pr(>F) 
Body.mass         330046    1    13.6030   0.006146 **
Gosner              5630    1     0.2321   0.642908 
Body.mass:Gosner   17736    1     0.7310   0.417423 
Residuals 194102 8 
---
Signif. codes: 0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1

Suggests interaction is not significant, i.e., the slopes are not different.

We can then proceed to check to see if the intercepts are different, now that we’ve confirmed no significant difference in slope.

R code for ANCOVA as additive model

tadpole.2 <- lm(VO2 ~ Body.mass + Gosner, data=example.Tadpole)
summary(tadpole.2)
Anova(tadpole.2, type="II")

Output

> summary(tadpole.2)

Call:
lm(formula = VO2 ~ Body.mass + Gosner, data = example.Tadpole)

Residuals:
    Min       1Q    Median      3Q      Max 
-163.12  -125.53    -20.27   83.71   228.56

Coefficients:
                   Estimate   Std. Error     t value    Pr(>|t|) 
(Intercept)         -595.37       239.87      -2.482     0.03487 * 
Body.mass            431.20       115.15       3.745     0.00459 **
Gosner[T.II]          64.96       132.83       0.489     0.63648 
---
Signif. codes: 0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1

Residual standard error: 153.4 on 9 degrees of freedom
(1 observation deleted due to missingness)
Multiple R-squared: 0.8047, Adjusted R-squared: 0.7613 
F-statistic: 18.54 on 2 and 9 DF, p-value: 0.0006432

Anova(tadpole.2, type="II")
Anova Table (Type II tests)

Response: VO2
              Sum Sq    Df    F value      Pr(>F) 
Body.mass     330046     1    14.0221   0.004593 **
Gosner          5630     1     0.2392   0.636482 
Residuals     211839     9 
---
Signif. codes: 0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1

Instead of the interaction, or as added, try a nested model, with body mass nested within stage.

tadpole.3 <- lm(VO2 ~ Body.mass/Gosner, data=example.Tadpole)

summary(tadpole.3)

Call:
lm(formula = VO2 ~ Body.mass/Gosner, data = example.Tadpole)

Residuals:
    Min       1Q    Median      3Q      Max 
-168.66  -131.14    -20.28   90.33   225.36

Coefficients:
                 Estimate      Std. Error    t value Pr(>|t|) 
(Intercept)           -575.10      319.51     -1.800   0.1054 
Body.mass              423.65      162.50      2.607   0.0284 *
Body.mass:Gosner[T.II]  21.95       63.73      0.344   0.7384 
---
Signif. codes: 0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1

Residual standard error: 154.4 on 9 degrees of freedom
(1 observation deleted due to missingness)
Multiple R-squared: 0.8021, Adjusted R-squared: 0.7581 
F-statistic: 18.24 on 2 and 9 DF, p-value: 0.0006823

Gets the true coefficient (nested lm() version).

The two test different hypotheses.

lm(VO2 ~ Body.mass * Gosner) tests whether or not the regression has a nonzero slope.

lm(VO2 ~ Body.mass */Gosner) test whether or not the slopes and intercepts from different factors are statistically significant.

Questions

1. Write up three learning outcomes for this page. Hint: Point your favorite generative AI to this page and ask for help.
2. An OLS approach was used for the analysis of tadpole oxygen consumption body mass. Consider the RMA approach — would that be a more appropriate regression model? Explain why or why not.
3. Consider an experiment in which you plan to administer a treatment that has a carry-over effect. For example, Compare and contrast “crossed” and “nested” designs.
4. True or False. The nested design option for the ANCOVA assumes the slopes for the two groups of tadpoles for the regression line of by are equal. Explain your choice.
5. Metabolic rates like oxygen consumption over time are well-known examples of allometric relationships. That is, many biological variables (eg, is related as aM^b, where M is body mass, slope b is scaling exponent), and best evaluated on log-log scale. Repeat the analysis above on log₁₀-transformed and for
   • crossed design (eg, tadpole.1 model)
   • added design (eg, tadpole.2 model)
   • nested design (eg, tadpole.3 model)
6. Create the plot and add the fitted lines from crossed design to the plot.

lgVO2 <- log10(VO2)

lm(log10(VO2) ~ log10(Body.mass ), data=example.Tadpole)

Quiz Chapter 17.6

ANCOVA - analysis of covariance

Start Quiz

Question

Confirm

✕

✓

Your answer:

Correct answer:

Next

Please wait..

↺

You got {{SCORE_CORRECT}} out of {{SCORE_TOTAL}}

Your Answers

Data sets

Oxygen consumption, , of Anuran tadpoles, dataset=example.Tadpole

Gosner refers to Gosner (1960), who developed a criteria for judging metamorphosis staging.

Chapter 17 contents

• Introduction
• Simple Linear Regression
• Relationship between the slope and the correlation
• Estimation of linear regression coefficients
• OLS, RMA, and smoothing functions
• Testing regression coefficients
• ANCOVA – Analysis of covariance
• Regression model fit
• Assumptions and model diagnostics for Simple Linear Regression
• References and suggested readings (Ch17 & 18)

18.1 – Multiple Linear Regression

Introduction

Last time we introduced simple linear regression:

• one independent X variable
• one dependent Y variable.

The linear relationship between Y and X was estimated by the method of Ordinary Least Squares (OLS). OLS minimizes the sum of squared distances between the observed responses, and responses predicted by the line. Simple linear regression is analogous to our one-way ANOVA — one outcome or response variable and one factor or predictor variable (Chapter 12.2).

But the world is complicated and so, our one-way ANOVA was extended to the more general case of two or more predictor (factor) variables (Chapter 14). As you might have guessed by now, we can extend simple regression to include more than one predictor variable. In fact, combining ANOVA and regression gives you the general linear model! And, you should not be surprised that statistics has extended this logic to include not only multiple predictor variables, but also multiple response variables. Multiple response variables falls into a category of statistics called multivariate statistics.

Like multi-way ANOVA, multiple regression is the extension of simple linear regression from one independent predictor variable to include two or more predictors. The benefit of this extension is obvious — our models gain realism. All else being equal, the more predictors, the better the model will be at describing and/or predicting the response. Things are not all equal, of course, and we’ll consider two complications of this basic premise, that more predictors are best; in some cases they are not.

However, before discussing the exceptions or even the complications of a multiple linear regression model, we begin by obtaining estimates of the full model, then introduce aspects of how to evaluate the model. We also introduce  and whether a reduced model may be the preferred model.

R code

Multiple regression is easy to do in Rcmdr — recall that we used the general linear model function, lm(), to analyze one-way ANOVA and simple linear regression. In R Commander, we access lm() by

Rcmdr: Statistics → Fit model → Linear model

You may, however, access linear regression through R Commander

We use the same general linear model function for cases of multi-way ANOVA and for multiple regression problems. Simply enter more than one ratio-scale predictor variable and boom!

You now have yourself a multiple regression. You would then proceed to generate the ANOVA table for hypothesis testing

Rcmdr: Models → Hypothesis testing → ANOVA tables

From the output of the regression command, estimates of the coefficients along with standard errors for the estimate and results of t-tests for each coefficient against the respective null hypotheses for each coefficient are also provided. In our discussion of simple linear regression we introduced the components: the intercept, the slope, as well as the concept of model fit, as evidenced by R², the coefficient of determination. These components exist for the multiple regression problem, too, but now we call the slopes partial regression slopes because there are more than one.

Our full multiple regression model becomes

where the coefficients β₁, β₂, … β_n are the partial regression slopes and β₀ is the Y-intercept for a model with 1 – n predictor variables. Each coefficient has a null hypothesis, each has a standard error, and therefore, each coefficient can be tested by t-test.

Now, regression, like ANOVA, is an enormous subject and we cannot do it justice in the few days we will devote to it. We can, however, walk you through a fairly typical example. I’ve posted a small data set diabetesCholStatin at the end of this page. Scroll down or click here. View the data set and complete your basic data exploration routine: make scatterplots and box plots.  We think (predict) that body size and drug dose cause variation in serum cholesterol levels in adult men. But do both predict cholesterol levels?

Selecting the best model

We have two predictor variables, and we can start to see whether none, one, or both of the predictors contribute to differences in cholesterol levels. In this case, both contribute significantly. The power of multiple regression approaches is that it provides a simultaneous test of a model which may have many explanatory variables deemed appropriate to describe a particular response. More generally, it is sometimes advisable to think more philosophically about how to select a best model.

In model selection, some would invoke Occam’s razor — given a set of explanations, the simplest should be selected — to justify seeking simpler models. There are a number of approaches (forward selection, backward selection, or stepwise selection), and the whole effort of deciding among competing models is complicated with a number of different assumptions, strengths and weaknesses. I refer you to the discussion below, which of course is just a very brief introduction to a very large subject in (bio)statistics!

Let’s get the full regression model

The statistical model is

As written in R format, our model is ChLDL ~ BMI + Dose + Statin.

We can now proceed in R Commander to fit the model.

Rmdr: Statistics → Fit models → Linear model

How the model is inputted into linear model menu is shown in Figure 1.

Figure 1. Screenshot of Rcmdr linear model menu with our model elements in place.

The output

summary(LinearModel.1)
Call:
lm(formula = ChLDL ~ BMI + Dose + Statin, data = cholStatins)

Residuals:
    Min      1Q  Median     3Q    Max
-3.7756 -0.5147 -0.0449 0.5038 4.3821

Coefficients:
                      Estimate Std. Error t value Pr(>|t|)
(Intercept)           1.016715   1.178430   0.863 0.39041
BMI                   0.058078   0.047012   1.235 0.21970
Dose                 -0.014197   0.004829  -2.940 0.00411 **
Statin[Statin2]       0.514526   0.262127   1.963 0.05255 .
---
Signif. codes: 0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1

Residual standard error: 1.31 on 96 degrees of freedom
Multiple R-squared: 0.1231, Adjusted R-squared: 0.09565
F-statistic: 4.49 on 3 and 96 DF, p-value: 0.005407

Question. What are the estimates of the model coefficients (rounded)?

• = intercept = 1.017
• = slope for variable BMI = 0.058
• = slope for variable Dose = -0.014
• = slope for variable Statin = -0.515

Question. Which of the three coefficients were statistically different from their null hypothesis (p-value < 5%)?

Answer: Only coefficient was judged statistically significant at the Type I error level of 5% (p = 0.0041).  Of the four null hypotheses we have for the coefficients:

• = 0
• = 0
• = 0; we only reject the null hypothesis for Dose coefficient.
• = 0)

The important take-home concept is about the lack of a direct relationship between the magnitude of the estimate of the coefficient and the likelihood that it will be statistically significant! In absolute value terms , but was not close to statistical significance (p = 0.220).

We generate a 2D scatterplot and include the regression lines (by group=Statin) to convey the relationship between at least one of the predictors (Fig 2).

Figure 2. Scatter plot of predicted LDL against dose of a statin drug. Regression lines represent the different statin drugs (Statin1, Statin2).

Question. Based on the graph, can you explain why there will be no statistical differences between levels of the statin drug type, Statin1 (shown open circles) vs. Statin2 (shown closed red circles).

Because we have two predictors (BMI and Statin Dose), you may also elect to use a 3D-scatterplot. Here’s one possible result (Fig 3).

Figure 3. 3D plot of BMI and dose of Statin drugs on change in LDL levels (green Statin2, blue Statin1).

R code for Figure 3.

Graph made in Rcmdr: Graphs → 3D Graph → 3D scatterplot …

scatter3d(ChLDL~BMI+Dose|Statin, data=diabetesCholStatin, fit="linear", 
residuals=TRUE, parallel=FALSE, bg="white", axis.scales=TRUE, grid=TRUE, 
ellipsoid=FALSE)

Figure 4. An example of a possible interactive 3D plot; the file embedded in this page is not interactive, just an animation.

Diagnostic plots

While visualization concerns are important, let’s return to the statistics. All evaluations of regression equations should involve an inspection of the residuals. Inspection of the residuals allows you to decide if the regression fits the data; if the fit is adequate, you then proceed to evaluate the statistical significance of the coefficients.

The default diagnostic plots (Fig 5) R provides are available from

Rcmdr: Models → Graphs → Basic diagnostics plots

Four plots are returned

Figure 5. R’s default regression diagnostic plots.

Each of these diagnostic plots in Figure 4 gives you clues about the model fit.

1. Plot of residuals vs. fitted helps you identify patterns in the residuals
2. Normal Q-Q plot helps you to see if the residuals are approximately normally distributed
3. Scale-location plot provides a view of the spread of the residuals
4. The residuals vs. leverage plot allows you to identify influential data points.

We introduced these plots in Chapter 17.8 when we discussed fit of simple linear model to data. My conclusion? No obvious trend in residuals, so linear regression is a fit to the data; data not normally distributed Q-Q plot shows S-shape.

Interpreting the diagnostic plots for this problem

The “Normal Q-Q” plot allows us to view our residuals against a normal distribution (the dotted line). Our residuals do no show an ideal distribution: low for the first quartile, about on the line for intermediate values, then high for the 3rd and 4th quartile residuals. If the data were bivariate normal we would see the data fall along a straight line. The “S-shape” suggests log-transformation of the response and or one or more of the predictor variables.

There also seems to be a pattern in residuals vs the predicted (fitted) values. There is a trend of increasing residuals as cholesterol levels increase, which is particularly evident in the “scale-location” plot. Residuals tended to be positive at low and high doses, but negative at intermediate doses. This suggests that the relationship between predictors and cholesterol levels may not be linear, and it demonstrates what statisticians refer to as a monotonic spread of residuals.

The last diagnostic plot looks for individual points that influence, change, or “leverage” the regression — in other words, if a point is removed, does the general pattern change? If so, then the point had “leverage” and thus we need to decide whether or not to include the datum. diagnostic plots Cook’s distance is a measure of the influence of a point in regression. Points with large Cook’s distance values warrant additional checking.

The multicollinearity problem

Statistical model building is a balancing act by the statistician. While simpler models may be easier to interpret and, perhaps, to use, it is a basic truism that the more predictor variables the model includes, the more realistic the statistical model. However, each additional parameter that is added to the statistical model must be independent of all other parameters already in the model. To the extent that this assumption is violated, the problem is termed multicollinearity. If predictor variables are highly correlated, then they are essentially just linear combinations and do not provide independent evidence. For example, one would naturally not include two core body temperature variables in a statistical model on basal metabolic rate, one in degrees Fahrenheit and the other in degrees Celsius, because  it is a simple linear conversion between the two units. This would be an example of structural collinearity: the collinearity is because of misspecification of the model variables. In contrast, collinearity among predictor variables may because the data are themselves correlated. For example, if multiple measures of body size are included: weight, height, length of arm, etc., then we would expect these to be correlated, ie, data multicollinearity.

Collinearity in statistical models may have a number of undesirable effects on a multiple regression model. These include

• estimates of coefficients not stable: with collinearity, values of coefficients depend on other variables in the model; if collinear predictors, then the assumption of independent predictor variables is violated.
• precision of the estimates decreases (standard error of estimates increase).
• statistical power decreases.
• p-values for individual coefficients not trust worthy.

Tolerance and Variance Inflation Factor

Absence of multicollinearity is important assumption of multiple regression. A partial test is to calculate product moment correlations among predictor variables. For example, when we calculate the correlation between BMI and Dose for our model, we get r = 0.101 (P = 0.3186), and therefore would tentatively conclude that there was little correlation between our predictor variables.

A number of diagnostic statistics have been developed to test for multicollinearity. Tolerance for a particular independent variable (Xi) is defined as 1 minus the proportion of variance it shares with the other independent variables in the regression analysis (1 − R²_i) (O’Brien 2007). Tolerance reports the proportion of total variance explained by adding the X_i–^th predictor variable that is unrelated to the other variables in the model. A small value for tolerance indicates multicollinearity — and that the predictor variable is nearly a perfect combination (linear) of the variables already in the model and therefore should be omitted from the model. Because tolerance is defined in relation to the coefficient of determination you can interpret a tolerance score as the unique variance accounted for by a predictor variable.

A second, related diagnostic of multicollinearity is called the Variance Inflation Factor, VIF. VIF is the inverse of tolerance.

VIF shows how much of the variance of a regression coefficient is increased because of collinearity with the other predictor variables in the model. VIF is easy to interpret: a tolerance of 0.01 has a VIF of 100; a tolerance of 0.1 has a VIF of 10; a tolerance of 0.5 has a VIF of 2, and so on. Thus, small values of tolerance and large values of VIF are taken as evidence of multicollinearity.

Rcmdr: Models → Numerical diagnostics → Variation-inflation factors

vif(RegModel.2)
      BMI     Dose 
1.010256 1.010256

A rule of thumb is that if VIF is greater than 5 then there is multicollinearity; with VIF values close to one we would conclude, like our results from the partial correlation estimate above, that there is little evidence for a problem of collinearity between the two predictor variables. They can therefore remain in the model.

Solutions for multicollinearity

If there is substantial multicollinearity then you cannot simply trust the estimates of the coefficients. Assuming that there hasn’t been some kind of coding error on your part, then you may need to find a solution. One solution is to drop one of the predictor variables and redo the regression model. Another option is to run what is called a Principal Components Regression (PCR).  One takes the predictor variables and runs a Principal Component Analysis to reduce the number of variables, then the regression is run on the PCA components. By definition, the PCA components are independent of each other. Another option is to use ridge regression approach.

Like any diagnostic rule, however, one should not blindly apply a rule of thumb. A VIF of 10 or more may indicate multicollinearity, but it does not necessarily lead to the conclusion that the linear regression model requires that the researcher reduce the number of predictor variables or analyze the problem using a different statistical method to address multicollinearity as the sole criteria of a poor statistical model. Rather, the researcher needs to address all of the other issues about model and parameter estimate stability, including sample size. Unless the collinearity is extreme (like a correlation of 1.0 between predictor variables!), larger sample sizes alone will work in favor of better model stability (by lowering the sample error) (O’Brien 2007).

Questions

1. Write up three learning outcomes for this page. Hint: Point your favorite generative AI to this page and ask for help.
2. Please explain why the magnitude of the slope is not the key to statistical significance of a slope? Hint: look at the equation of the t-test for statistical significance of the slope.
3. Consider the following scenario. A researcher measures his subjects multiple times for blood pressure over several weeks, then plots all of the values over time. In all, the data set consists of thousands of readings. He then proceeds to develop a model to explain blood pressure changes over time. What kind of collinearity is present in his data set? Explain your choice.
4. We noted that Dose could be viewed as categorical variable. Convert Dose to factor variable (fDose) and redo the linear model. Compare the summary output and discuss the additional coefficients.
   • Use Rcmdr: Data → Manage variables in active data set →  Convert numeric Variables to Factors to create a new factor variable fDose. It’s ok to use the numbers as factor levels.
5. We flagged the change in LDL as likely to be not normally distributed. Create a log10-transformed variable for ChLDL and perform the multiple regression again.
   1. Write the new statistical model
   2. Obtain the regression coefficients — are they statistically significant?
   3. Run basic diagnostic plots and evaluate for

Quiz Chapter 18.1

Multiple Linear Regression

Start Quiz

Question

Confirm

✕

✓

Your answer:

Correct answer:

Next

Please wait..

↺

You got {{SCORE_CORRECT}} out of {{SCORE_TOTAL}}

Your Answers

Data set

Chapter 18 contents

• Introduction
• Multiple linear regression
• Nonlinear regression
• Logistic regression
• Generalized Linear Squares
• Selecting the best model
• Compare two linear models
• References and suggested readings (Ch17 & 18)

16 – Correlation, Similarity, and Distance

Introduction

We continue with our discussion and introduction of inferential statistics. Recall that as we analyze a data set, we generally want to begin by describing it (central tendency, measures of variability), and we also want to plot the data. To begin our introduction to correlation and regression, first we describe how to produce graphs to help show linear association or in some cases, cause and effect — the latter perhaps the primary reason for using regression.

Graphical representation

The previous statistical procedures we have examined have used one or more categorical or qualitative variables (Chapter 3). For example,

1. Chi-Square Analyses: variables are all categorical, including the response variable (Chapter 9).
2. T-tests: one categorical (Factor) variable and one (Dependent, Outcome, Response) variable that was continuous or interval scale (Chapter 8.5, 10).
3. ANOVA Analyses: one or more variables are categorical (Factors, the independent variables) and one (Dependent, Outcome, Response) variable that was continuous or interval scale (Chapter 12, 14).

The convention in graphing ANOVA (or Chi-Square) is to use the Factor or Independent variables as the
X-axis and to have the dependent variable (Response) as the Y-axis. We called these bar charts (Chapter 4.1).

Figure 1. Bar chart with error bars.

Box plots (Chapter 4.3) are also useful, and perhaps the preferred choice to display this type of comparison (one involving groups) (Fig 2).

Figure 2. Box plots.

In correlation (and regression) analyses we will have two or more continuous or interval scale variables. To show relationships among continuous variables, a scatter plot, also called an X-Y plot, works well (Chapter 4.5).

In correlation, no causation is implied, so either variable can be placed on the X-axis. The convention of graphing in regression is to place the independent variable as the X-axis and the dependent variable as the Y-axis (Fig 3). Another consideration: if one variable is considered fixed and the other random, then the fixed variable would be assigned to the horizontal axis.

Figure 3. Scatterplot with groups

To produce a scatterplot (also called an X-Y plot) in Rcmdr, select Graph → Plot → and select the Y and X variables. Use a combination of Options, Frame, and Edit Attributes selections to modify the default graph.

Suggested homework

Homework 10: Correlation.

Quizzes in this chapter

A total of 64 questions among the several subchapters, a mix of true or false and multiple choice question format.

Chapter 16 contents

• Introduction
• Product moment correlation
• Causation and Partial correlation
• Data aggregation and correlation
• Spearman and other correlations
• Instrument reliability and validity
• Similarity and Distance
• References and suggested readings

4.5 – Scatter plots

Introduction

Scatter plots, also called scatter diagrams, scatterplots, or XY plots, display associations between two quantitative, ratio-scaled variables. Each point in the graph is identified by two values: its X value and its Y value. The horizontal axis is used to display the dispersion of the X variable, while the vertical axis displays the dispersion of the Y variable.

The graphs we just looked at with Tufte’s examples Anscombe’s quartet data were scatter plots (Chapter 4 – How to report statistics).

Here’s another example of a scatter plot, data from Francis Galton, data contained in the R package HistData.

Figure 1. Scatterplot of mid-parent (vertical axis) and their adult children’s (horizontal axis) height, in inches. data from Galton’s 1885 paper, “Regression towards mediocrity in hereditary stature.” The red line is the linear regression fitted line, or “trend” line, which is interpreted in this case as the heritability of height.

The commands I used to make the plot in Figure 1 were

library(HistData)
data(GaltonFamilies, package="HistData")
attach(GaltonFamilies)

plot(childHeight~midparentHeight, xlab="Mid-parent height", ylab="Mean adult children's height", xlim=c(55, 80), ylim=c(55,80), cex=0.8, cex.axis=1.2, cex.lab=1.3, pch=c(19), data=GaltonFamilies)
abline(lm(childHeight~midparentHeight), col="red", lwd=2)

I forced the plot function to use the same range of values, set by providing values for xlim and ylim; the default values of the plot command picks a range of data that fits each variable independently. Thus, the default X axis values ranged from 64 to 76 and the Y variable values ranged from 55 to 80. This has the effect of shifting the data, reducing the amount of white space, which a naïve reading of Tufte would suggest is a good idea, but at the expense of allowing the reader to see what would be the main point of the graph: that the children are, on average, shorter than the parents, mean height = 67 vs. 69 inches, respectively. Therefore, Galton’s title begins with the word “regression,” as in the definition of regression as a “return to a former … state” (Oxford Dictionary).

For completeness, cex sets the size of the points (default = 1), and therefore cex.axis and cex.lab apply size changes to the axes and labels, respectively; pch refers to the graph elements or plotting characters, further discussed below (see Fig 8); lm() is a call to the linear model function; col refers to color.

Figure 2 shows the same plot, but without attention to the axis scales, and, more in keeping with Tufte’s principle of maximize data, minimize white space.

Figure 2. Same plot as Figure 1, but with default settings for axis scales.

Take a moment to compare the graphs in Figure 1 and 2. Setting the scales equal allows you to see that the mid-parent heights were less variable, between 65 and 75 inches, than the mean children height, which ranged from 55 to 80 inches.

And another example, Figure 3. This plot is from the ggplot2() function and was generated from within R Commander’s KMggplot2 plug-in.

Figure 3. Finishing times in minutes of 1278 runners by age and gender at the 2013 Jamba Juice Banana 5K in Honolulu, Hawaiʻi. Loess smoothing functions by groups of female (red) and male (blue) runners are plotted along with 95% confidence intervals.

Figure 3 is a busy plot. Because there were so many data points, it is challenging to view any discernible pattern, unlike Figure 1 and 2 plots, which featured less data. Use of the loess smoothing function, a transformation of the data to reduce data “noise” to reveal a continuous function, helps reveal patterns in the data:

1. across most ages, men completed 5K faster than did females and
2. there was an inverse, nonlinear association between runner’s age and time to complete the 5K race.

Take a look at the X-axis. Some runners ages were reported as less than 5 years old (trace the points down to the axis to confirm), and yet many of these youngsters were completing the 5K race in less than 30 minutes. That’s 6-minute mile pace! What might be some explanations for how pre-schoolers could be running so fast?

Design criteria

As in all plotting, maximize information to background. Keep white space minimal and avoid distorting relationships. Some things to consider:

1. keep axes same length
2. do not connect the dots UNLESS you have a continuous function
3. do not draw a trend line UNLESS you are implying causation

Scatter plots in R

We have many options in R to generate scatter plots. We have already demonstrated use of plot() to make scatter plots. Here we introduce how to generate the plot in R Commander.

Rcmdr: Graphs → Scatterplot…

Rcmdr uses the scatterplot function from the car package. In recent versions of R Commander the available options for the scatterplot command are divided into two menu tabs, Data and Options, shown in Figure 4 and Figure 5.

Figure 4. First menu popup in R Commander Scatterplot command, Rcmdr ver. 2.2-3.

Select X and Y variables, choose Plot by groups if multiple grounds are included, eg, male, female, then click Options tab to complete.

Figure 5. Second menu popup in R Commander scatterplot command., Rcmdr ver. 2.2-3

Set graph options including axes labels and size of the points.

Here is another example using the default settings in scatterplot() function in the car package, now the default scatter plot command via R Commander (Fig. 4), along with the same graph, but modified to improve the look and usefulness of the graph (Fig. 6). The data set was Puromycin in the package datasets.

Figure 6. Default scatterplot, package car, from R Commander, version 2.2-4.

Grid lines in graphs should be avoided unless you intend to draw attention to values of particular data points. I prefer to position the figure legend within the frame of the graph, eg, the open are at the bottom right of the graph. Modified graph shown in Figure 7.

Figure 7. Modified scatterplot, same data from Figure 6

R commands used to make the scatter plot in Figure 7 were

scatterplot(rate~conc|state, col=c("blue", "red"), cex=1.5, pch=c(19,19), 
bty="n", reg=FALSE, grid=FALSE, legend.coords="bottomright")

A comment about graph elements in R

In some ways R is too rich in options for making graphs. There are the plot functions in the base package, there’s lattice and ggplot2 which provide many options for graphics, and more. The advice is to start slowly, for example taking advantage of the Figure 8 displays R’s plotting characters and the number you would invoke to retrieve that plotting character.

Figure 8. R plotting characters pch = 1 – 25 along with examples of color.

colors()

  [1] "white"                "aliceblue"            "antiquewhite" 

[655] "yellow3"              "yellow4"              "yellowgreen"  

There’s a lot more to R plotting. For example, you are not limited to just 25 possible characters. R can print any of the ASCII characters 32:127 or from the extended ASCII code 128:255. See Wikipedia to see the listing of ASCII characters.

In addition to pch, you can change the size of the plotting character with cex, color of the character with col for lines or outlines, and bg to change background color. lwd is used to set the character’s border thickness.

Here’s the R code used to generate the graph in Figure 8. Remember, any line beginning with # is a comment line, not an R command.

#create a vector with 26 numbers, from 0 to 25
stuff <-c(0:25)
plot(stuff, pch=c(32:58), cex = 2.5, col = c(1:26), 'xlab' = "pch number", 'ylab' = "Plotting character number")

Is it “scatter plot” or “scatterplot”?

Spelling matters, of course, and yet there are many words for which the correct spelling seems to be like “beauty,” it is “in the eye of the beholder” (Molly Bawn, 1878, by Margaret Hungerford). Scatter plot is one of these — is it one word or two, or is it something else entirely?

Scatter plot is one of these terms: you’ll find it spelled as “scatterplot” or as “scatter plot,” in the dictionary (eg, Oxford English dictionary), with no guidance to choose between them. And I’m not just talking about the differences between British and American English spelling traditions. 

Thus, in these situations as an author, you can turn to which of the spellings is in common use. I first looked at some of the statistics books on my shelves. I selected 14 (bio)statistics textbooks and checked the index and if present, chapters on graphics for term usage.

Table 1. Frequency of use of different terms for scatter plot in 14 (bio)statistics books currently on Mike’s shelves.

Not much help, basically, it is a tie between “scatter plot” and “scatterplot.”

Next, I searched six journals for the interval 1990 – 2016 for use of these terms. Results presented in Table 6 along with journal impact factor for 2014 and number of issues (Table 2).

Table 2. Impact factor and number of issues 1990 – 2016 for six science journals.

My methods? I used the journal’s online search functions for the various usage for scatter plot and the results are shown in Figure 9.

Figure 9. Usage of terms for X Y plots in research articles normalized to number of issues in six journals between 1990 and 2016.

The journals have different numbers of articles; I partially corrected for this by calculating the ratio number of articles with one of the terms divided by the number of issues for the interval 1990 – 2016. It would have been better to count all of the articles, but even I found that to be an excessive effort given the point I’m trying to make here.

Not much help there, although we can see a trend favoring “scatter plot” over any of the other options.

And finally, to completely work over the issue I present results from use of Google’s Ngram Viewer. Ngram Viewer allows you to search words in all of the texts that Google’s folks have scanned into digital form. I searched on the terms in texts between 1950 and 2015, and results are displayed in Figure 10 and Figure 11.

Figure 10. Results from Ngram Viewer for American English, “scatterplot” (blue), “scatter plot” (red), “scatter diagram” (green), “scattergram” (orange), and “XY plot” (purple).

And the same plot, but this time for British sources

Figure 11. Results from Ngram Viewer for British English. See Figure 10 for key.

Conclusion? It looks like “scatterplot” (blue line) is the preferred usage, but it is close. Except for “scattergram” and “XY plot,” which, apparently, are rarely used. After all of this, it looks like you’re free to make your choice between “scatterplot” or “scatter plot.” I will continue to use “scatter plot.”

Bland-Altman plot

Also known as Tukey mean-difference plot, the Bland-Altman plot is used to describe agreement between two ratio scale variables (Bland and Altman 1986, Giavarina 2015), for example agreement between two different methods used to measure the same samples.

Consider use of imageJ by two different observers to record number of pixels of a unit measure (1 cm) on a series of digital images — there’s subjectivity in drawing the lines (where to start, where to end) — do they agree?  Data set below. blandr package, blandr.draw function.

Figure 12. Bland-Altman plot of 1 cm unit measure in pixel number by imageJ from digital images by two independent observers. Purple central region is 95% CI. Lower and upper dashed horizontal lines represent bounds of “acceptable” agreement.

blandr.draw(Obs1, Obs2)

The plot makes it easy to identify questionable points, for example, the one point in upper right quadrant looks suspect.

Visualizing large data sets

Since the early 2000s, size of data sets in biology have increased dramatically, leaving many statistical graphics solutions ill-equipped to tell stories. The data scientists involved with the genomics and health care communities have introduced a number of useful visualizations. We introduce just two, the volcano plot (Fig 13) and the Manhattan-style plot (Fig 14).

Volcano plot

The volcano plot (Fig 13) is a common way to show events that differ between two groups of subjects (eg, p-values), and is common often provides to help visualize gene expression studies of an exposed group vs a control group (eg, fold changes). For the volcano plot, fold changes are reflected on horizontal axis, indicating how much the gene expression level has increased or decreased. The vertical axis typically represents the negative logarithm of the p-value (-logP), which indicates the statistical significance of the change.

Below is R code — with assistance from Gemini AI (Fall 2026) — for a set of five replicate simulated gene expression fold change data sets on stress test (NaCl, H₂O₂) on yeast grown on YPDA plates. Gene expression profiles were simulated for a set of  100 genes (oxidative pathway, DNA repair pathway); as noted, such studies typically include many more genes. 

Figure 13. Volcano plot, gene expression fold change on simulated stress test (NaCl, H₂O₂) yeast data sets, 100 genes (oxidative pathway, DNA repair pathway).

Inspecting the plots (Fig 13), we see a symmetric distribution: The overall distribution of genes is cleanly centered around a Log2 Fold Change (LFC) of zero. This mirrors a realistic transcriptomic profile where the vast majority of the background genome (shown in grey) remains stable, displaying only technical and biological noise. There is a distinct, asymmetrical shift toward positive fold changes among specific gene pathways (targeted activation). Genes belonging to the Oxidative Stress pathway (red) and DNA Repair pathway (blue) consistently surpass both the statistical threshold () and the biological threshold (). Additionally, we see evidence of synergistic effects: The simultaneous application of both stressors (“Both”) drives these pathways to their highest expression levels, pushing them into the top-right quadrant of the plots as highly significant, upregulated features.

R code — with assistance from Gemini AI (Fall 2026) — for yeast gene expression and volcano plots; graphs by base package only.

# ==============================================================================
# 1. SIMULATE 5 REASONABLE DATASETS WITH BIOLOGICAL & TECHNICAL NOISE
# ==============================================================================
set.seed(123) # New seed for realistic distribution

n_genes <- 100
n_datasets <- 5
groups <- c("Control", "Salt", "H2O2", "Both")
n_groups <- length(groups)

# Real-world experiment: each dataset has 3 biological replicates per group
n_reps_per_group <- 3

# Define gene pathways (15 oxidative stress, 15 DNA repair, 70 background/unchanged)
gene_ids <- paste0("Gene_", 1:n_genes)
pathways <- c(rep("Oxidative_Stress", 15), rep("DNA_Repair", 15), rep("Background", 70))

# Initialize a list to hold the 5 realistic datasets
datasets <- list()

for (d in 1:n_datasets) {

# 1. Establish a realistic base expression distribution across 100 genes
# Real data has high-expressing and low-expressing genes
base_expression <- rnorm(n_genes, mean = 7, sd = 2)

# Create a blank sample matrix: 100 genes x 12 samples (4 groups x 3 replicates)
sample_names <- paste0(rep(groups, each = n_reps_per_group), "_R", 1:n_reps_per_group)
expr_matrix <- matrix(0, nrow = n_genes, ncol = length(sample_names))
rownames(expr_matrix) <- gene_ids
colnames(expr_matrix) <- sample_names

# 2. Add realistic baseline biological variation to every sample
for (col_idx in 1:ncol(expr_matrix)) {
expr_matrix[, col_idx] <- base_expression + rnorm(n_genes, mean = 0, sd = 0.4)
}

# 3. Inject subtle, realistic treatment effects (not over-exaggerated)
for (i in 1:n_genes) {
# Every gene gets a tiny bit of random directional drift across groups (real-world noise)
drift <- rnorm(n_groups, mean = 0, sd = 0.15)

if (pathways[i] == "Oxidative_Stress") {
# Subtle upregulation in H2O2 and Both (mean fold change ~ 1.0 to 1.3)
drift[3] <- drift[3] + rnorm(1, mean = 1.1, sd = 0.2) # H2O2
drift[4] <- drift[4] + rnorm(1, mean = 1.3, sd = 0.2) # Both
} else if (pathways[i] == "DNA_Repair") {
# Subtle upregulation in Salt and Both (mean fold change ~ 0.8 to 1.2)
drift[2] <- drift[2] + rnorm(1, mean = 0.9, sd = 0.2) # Salt
drift[4] <- drift[4] + rnorm(1, mean = 1.2, sd = 0.2) # Both
}

# Apply the group drifts to the respective sample columns
for (g_idx in 1:n_groups) {
start_col <- (g_idx - 1) * n_reps_per_group + 1
end_col <- g_idx * n_reps_per_group
expr_matrix[i, start_col:end_col] <- expr_matrix[i, start_col:end_col] + drift[g_idx]
}
}

datasets[[d]] <- expr_matrix
}

# ==============================================================================
# 2. STATISTICAL ANALYSIS (Calculated directly from replicates)
# ==============================================================================
# We will compare the double stressor 'Both' vs 'Control' using real t-tests.
all_lfc <- matrix(0, nrow = n_genes, ncol = n_datasets)
all_pvals <- matrix(0, nrow = n_genes, ncol = n_datasets)
rownames(all_lfc) <- rownames(all_pvals) <- gene_ids

for (d in 1:n_datasets) {
mat <- datasets[[d]]

# Identify the columns for each group
control_cols <- grep("^Control_", colnames(mat))
both_cols <- grep("^Both_", colnames(mat))

for (i in 1:n_genes) {
control_vals <- mat[i, control_cols]
both_vals <- mat[i, both_cols]

# Real Log2 Fold Change (Difference of means since data is on log scale)
all_lfc[i, d] <- mean(both_vals) - mean(control_vals)

# Real Student's t-test
ttest_res <- t.test(both_vals, control_vals, var.equal = TRUE)
all_pvals[i, d] <- ttest_res$p.value
}
}

# ==============================================================================
# 3. GENERATE REALISTIC VOLCANO PLOTS
# ==============================================================================
par(mfrow = c(2, 3), mar = c(4.5, 4.5, 2.5, 1.5))

lfc_threshold <- 1.0 
p_threshold <- 0.05

for (d in 1:n_datasets) {
lfc <- all_lfc[, d]
neg_log10_p <- -log10(all_pvals[, d])

col_vector <- rep("grey75", n_genes)

is_sig_up <- (lfc >= lfc_threshold) & (all_pvals[, d] <= p_threshold)
is_sig_down <- (lfc <= -lfc_threshold) & (all_pvals[, d] <= p_threshold)

col_vector[is_sig_up & pathways == "Oxidative_Stress"] <- "firebrick2"
col_vector[is_sig_up & pathways == "DNA_Repair"] <- "dodgerblue2"
col_vector[(is_sig_up | is_sig_down) & col_vector == "grey75"] <- "grey30"

plot(lfc, neg_log10_p,
col = col_vector,
pch = 19,
cex = 0.9,
xlab = expression(Log ~ Fold ~ Change),
ylab = expression(-Log ~ p - value),
main = paste("Dataset Replicate", d),
xlim = c(-2.5, 2.5), 
ylim = c(0, 5), 
font.main = 1)

abline(v = c(-lfc_threshold, lfc_threshold), col = "grey50", lty = 3)
abline(h = -log10(p_threshold), col = "grey50", lty = 3)

# Set the legend.
if (d == 1) {
legend("topleft", 
legend = c("Oxidative Stress (Up)", "DNA Repair (Up)", "Other Significant", "Not Significant"),
col = c("firebrick2", "dodgerblue2", "grey30", "grey75"), 
pch = 19, 
cex = 0.6,
box.lty = 0,
bg = "transparent")
}
}

# Reset layout parameters
par(mfrow = c(1, 1))

End R code.

About the simulation

1. Symmetric, Balanced Center: Most of the 100 genes (the 70 background genes) are strictly driven by standard biological and technical noise. This distributes them evenly around 0 on the X-axis, meaning you will naturally see negative fold changes alongside positive ones.
2. Realistic Effect Scales: True upregulations are dialed down to max out around to Fold Changes, mirroring standard biological responses on YPDA plates.
3. True T-Tests: Instead of faking p-values, each simulated dataset now contains 3 biological samples per group. The p-values are calculated dynamically using t.test(), generating a mathematically sound distribution pattern on the Y-axis.

Manhattan plot

A Manhattan plot (Fig 14) is a type of scatter plot used to display the statistical significance of large sets of data points simultaneously. Some feature of the experiment, for example location in the genome or sequential test identifiers, is displayed on the horizontal axis. Like the volcano plot (Fig 13), the vertical axis displays the statistical significance (eg, -log₁₀(p-value)).

Figure 14. Manhattan plot, gene expression fold change on simulated stress test (NaCl, H2O2) yeast data sets, 100 genes (oxidative pathway, DNA repair pathway).

Looking over the plots (Fig 14) we see evidence of pathway clustering: By arranging the 100 genes sequentially across three chromosomes, we see distinct localized peaks rather than a random distribution of significant points. We also see chromosome hotspots: Chromosome I (Oxidative Stress genes) and Chromosome II (DNA Repair genes) display pronounced vertical clusters that reliably break through the red significance line across all five experimental replicates. This simulates a real-world genomic architecture where functionally related gene clusters or operon-like structures respond in unison to environmental shifts.

Conversely, Chromosome III (Background genes, ie, baseline Control) remained entirely flat and well below the significance threshold, validating that the observed stress responses are highly specific and not an artifact of global transcriptional runaway.

R code — with assistance from Gemini AI (Fall 2026) —  for Manhattan plot.

# ==============================================================================
# 1. ASSIGN GEOMETRIC CHROMOSOME POSITIONS TO SIMULATED GENES
# ==============================================================================
# We will use the exact same 5 datasets generated in the previous step.
# To plot a Manhattan plot, every gene needs a Chromosome and a Base Pair (BP) position.

# Assign chromosomes: 
# Chromosome 1: Genes 1-40 (Oxidative Stress pathway)
# Chromosome 2: Genes 41-80 (DNA Repair pathway)
# Chromosome 3: Genes 81-100 (Background/Control pathway)
chroms <- c(rep(1, 40), rep(2, 40), rep(3, 20))

# Assign incremental Base Pair positions along each chromosome
bp_positions <- c(seq(1000, 40000, length.out = 40), 
seq(1000, 40000, length.out = 40), 
seq(1000, 20000, length.out = 20))

# To make a continuous X-axis across multiple chromosomes, we calculate cumulative positions
chr_lengths <- c(45000, 45000, 25000)
chr_offsets <- c(0, cumsum(chr_lengths)[1:2])
coord_x <- bp_positions + chr_offsets[chroms]

# Calculate where the chromosome tick marks should sit on the X-axis (the midpoints)
chr_midpoints <- chr_offsets + (chr_lengths / 2)

# ==============================================================================
# 2. GENERATE MANHATTAN PLOTS FOR ALL 5 REPLICATES
# ==============================================================================
# Set up a clean 2x3 plotting grid to display all 5 replicates together
par(mfrow = c(2, 3), mar = c(4.5, 4.5, 2.5, 1.5))

# Standard scientific significance line (-log10(0.05) is ~1.3)
p_threshold <- 0.05
suggestive_line <- -log10(p_threshold)

for (d in 1:n_datasets) {
neg_log10_p <- -log10(all_pvals[, d])

# Alternating color scheme per chromosome to differentiate boundaries
# Chromosome 1 = Slate Blue, Chromosome 2 = Deep Blue, Chromosome 3 = Dark Grey
chr_colors <- c("lightslateblue", "steelblue4", "grey40")
col_vector <- chr_colors[chroms]

# Create the Manhattan plot grid
plot(coord_x, neg_log10_p,
col = col_vector,
pch = 19,
cex = 1.0,
xaxt = "n", # Suppress the standard numeric X-axis to draw custom Chromosome ticks
xlab = "Chromosome",
ylab = expression(-Log[10] ~ p - value),
main = paste("Dataset Replicate", d),
xlim = c(0, sum(chr_lengths)),
ylim = c(0, 5),
font.main = 1)

# Add the standard genome-wide significance horizontal threshold line
abline(h = suggestive_line, col = "firebrick3", lty = 2, lwd = 1.2)

# Draw the custom structural X-axis labeling the specific chromosomes
axis(1, at = chr_midpoints, labels = c("Chr I", "Chr II", "Chr III"))

# Add an explanatory legend to the first plot panel to map pathways to locations
if (d == 1) {
legend("topright", 
legend = c("Chr I (Oxidative Stress)", "Chr II (DNA Repair)", "Chr III (Background)"),
col = c("lightslateblue", "steelblue4", "grey40"), 
pch = 19, 
cex = 0.65,
box.lty = 0,
bg = "transparent")
}
}

# Reset layout parameters back to default single screen layout
par(mfrow = c(1, 1))

About the R code.

The Genomic Peaks: Because these simulated Oxidative Stress genes sit together on Chromosome I and DNA Repair genes sit on Chromosome II,  we see pronounced, beautiful vertical clusters extending upwards over the red significance line.

The Noise Flatline: Chromosome III hosts the background genes, creating a realistically flat landscape where points rarely cross the significance threshold.

Alternating Chromosome Tracks: Using the standard coordinate offset method, the data points stay organized left-to-right without overlapping.

Questions

1. Using our Comet assay data set (Table 1, Chapter 4.2), create scatter plots to show associations between tail length, tail percent, and olive moment. 
2. Explore different settings including size of points, amount of white area, and scale of the axes. Evaluate how these changes change the “story” told by the graph.

Quiz Chapter 4.5

Scatter plots

Start Quiz

Question

Confirm

✕

✓

Your answer:

Correct answer:

Next

Please wait..

↺

You got {{SCORE_CORRECT}} out of {{SCORE_TOTAL}}

Your Answers

Data sets

Number of pixels of 1 cm length from unit measure on ten digital images. Recorded by two independent observers, Obs1 and Obs2.

 Obs1, Obs2
171.026, 171.105
136.528, 138.521
148.084, 144.222
142.014, 140.057
150.213, 153.118
187.011, 195.092
168.760, 168.668
154.302 ,160.381
209.022, 209.876
240.067, 161.805

Gene expression

ccc

Chapter 4 contents

• Graphs and tables (How to report statistics)
• Bar (column) charts
• Histograms
• Box plot
• Mosaic plots
• Scatter plots
• Add a second Y axis
• Q-Q plot
• Ternary plots
• Heat maps
• Graph software
• References and suggested readings

3.3 – Measures of dispersion

More about summary statistics.

The statistics of data exploration involve calculating estimates of the middle, central tendency, and the variability, or dispersion, about the middle. Statistics about the middle were presented in the previous section, Chapter 3.1. Statistics about measures of dispersion, and how to calculate them in R, are presented in this page. Use of Z score to standardize or normalize scores is introduced. Statistical bias is also introduced. 

Describing the middle of the data gives your reader a sense of what was the typical observation for that variable. Next, your reader will want to know something about the variation about the middle value — what was the smallest value observed? What was the largest value observed? Were data widely scattered or clumped about the middle?

Measures of dispersion or variability are descriptive statistics used to answer these kinds of questions. Variability statistics are very important and we will use them throughout the course. A key descriptive statement of your data, how variable?

Examine the two figures below (Fig 1 and Fig 2): the two sample frequency distributions (put data into several groups, from low to high, and membership is counted for each group) have similar central tendencies (mean), but they have different degrees of variability (standard deviation).

Figure 1. A histogram which displays a sampling of data with a mean of 10 (arrow marks the spot) and standard deviation (sd) of 50 units.

Figure 2. A histogram which displays a sampling of data with the same mean of 10 (arrow marks the spot) displayed in Fig. 1, but with a smaller standard deviation (sd) of 5 units.

Clearly, knowing something about the middle of a data set is only part of the required information we need when we explore a data set; we need measures of dispersion as well. Provided certain assumptions about the frequency of observations hold, estimates of the middle (eg, mean, median) and the dispersion (eg, standard deviation) are adequate to describe the properties of any observed data set.

For measures of dispersion or variability, the most common statistics are: Range, Mean Deviation, Variance, Standard Deviation, and Coefficient of Variation.

The range.

The range is reported as one number, the difference between maximum and minimum values.

Arrange the data from low to high, subtract the minimum, X_min, from the maximum, X_max, value and that’s the range.

For example, the minimum height of a collection of fern trees might be 1 meter whereas the maximum height might be 2.2 meters. Therefore, the range is 1.2 meters (= 2.2 – 1).

The range is useful, but may be misleading (as index of true variability in population) if there is one or more exceptional “outlier” (one individual that has an exceptionally large or small value). I often just report X_min and X_max. That’s the way R does it, too.

If you recall, we used the following as an example data set.

x = c(4,4,4,4,5,5,6,6,6,7,7,8,8,8,15)

The command for range in R is simply range().

range(x)
[1] 4 15

There is apparently no universally accepted symbol for “range,” so we report the statistic as range = 15 – 4 = 11.

You may run into an interquartile range, which is abbreviated as IQR. It is a trimmed range, which means, like the trimmed mean, it will be robust to outlier observations. To calculate this statistic, divide the data into fourths, termed quartiles. For our example variable, x, we can get the quartiles in R

quantile(x)
  0%  25%  50%  75% 100% 
 4.0  4.5  6.0  7.5 15.0 

Thus, we see that 25% of the observations are 4.5 (or less), the second quartile is the median, and 75% of the observations are less than 7.5. The IQR is the difference between the first quartile, Q1, and the third quartile, Q3.

The command for obtaining the IQR in R is simply IQR(). Yes. You have to capitalize “IQR” — R is case sensitive; that means IqR is not the same as IQR or iqr.

IQR(x)
[1] 3.5

and we report the statistic as IQR = 3.5. Thus, 75% of the observations are within 3.5 points of the median. The IQR is used in box plots (Chapter 4). Quartiles are special cases of the more general quantile; quantiles divide up the range of values into groups with equal probabilities. For quartiles, four groups at 25% intervals. Other common quantiles include deciles (ten equal groups, 10% intervals) and percentiles (100 groups, 1% intervals).

The mean deviation.

Subtract each observation from the sample mean; each  is called a deviate: some observations will be positive (greater than ) and some will be negative (less than ).

Take the absolute value of the deviation and then add up the absolute values of the deviations from the mean. At the end, divide by the sample size, n. Large values for this statistic implies that much of the data is spread out, far from the mean. Small values in turn imply that each observation is close to the mean.

In R, we can get the mean deviation with the mad() function. At the R prompt, then

mad(x)
[1] 2.9652

Population variance and population standard deviation.

The population variance is appropriate for describing variability about the middle for a census. Again, a census implies every member of the population was measured. Equation of the population variance is

The population standard deviation also describes the variability about the middle, but has the advantage of being in the same units as the quantity (ie, no “squared” term). Equation of the population standard deviation is

Sample variance and sample standard deviation.

The above statements about the population variance and standard deviation hold for the sample statistics. Equation of the sample variance

and the equation of the sample standard deviation

Of course, instead of taking the square-root of the sample variance, the sample standard deviation could be calculated directly.

*** QuickLaTeX cannot compile formula:
\begin{align*}s = \sqrt{\frac{\sum \left ( X_{i} - \bar{X} \right )^2}{n-1}\end{align*}

*** Error message:
File ended while scanning use of \align*.
Emergency stop.

Statistical bias is an important concept in its own right; bias is a problem because it refers to a situation in which an estimator consistently returns a value different from the population parameter for which it is intended to estimate. Thus, the sample mean is said to be an unbiased estimator of the population mean. Here, bias means that the formula will give you a good estimate of the value. This turns out not to be the case for the sample variance if you divide by n instead of n - 1. Now, for very large values of n, this is not much of an issue, but it shows up when n is small. In R you get what you ask for — if you ask for the sample standard deviation, the software will return the correct value; calculators, go to watch out for this — not all of them are good at communicating which standard deviation they are calculating, the population or the sample standard deviation.

Winsorized variances.

Like the trimmed mean and winsorized mean (Ch3.2), we may need to construct a robust estimate of variability less sensitive to extreme outliers.  Winsorized refers to procedures to replace extreme values in the sample with a smaller value. As noted in Chapter 3.2, we chose the level ahead of time, eg, 90%. Winsorized values then  The winsorized variance is just the sample variance of the winsorized values.  winvar() from the WRS2 package.

Making use of the sample standard deviation.

Given an estimate of the mean and an estimate of the standard deviation, one can quickly determine the kinds of observations made and how frequently they are expected to be found in a sample from a population (assuming a particular population distribution). For example, it is common in journal articles for authors to provide a table of summary statistics like the mean and standard deviation to describe characteristics of the study population (aka the reference population), or samples of subjects drawn from the study population (aka the sample population). The CDC provides reports of attributes of for a sample of adults (more than forty thousand) from the USA population (Fryar et al 2018). Table 1 shows a sample of results for height, weight, and waist circumference for men aged 20 – 39 years.  

Table 1. Summary statistics mean ( standard deviation) of height, weight, and waist circumference of 20-39 year old men USA.

We introduced the normal deviate as a way to normalize scores, and which we we will use extensively in our discussion of the normal distribution in Chapter 6.7, as a way to standardize a sample distribution, assuming a normal curve.

For data that are normally distributed, the standard deviation can be used to tell us a lot about the variability of the data:

​ 62.26% of the data will lie between of

​ 95.46% of the data will lie between of  

​ 99.0% of the data will lie between of

This is known as the empirical rule, where 68% of the observations will be within one standard deviation, 95% of observations will be within two standard deviations, and 99% of observations will be within three standard deviations. 

For example, men aged 20 years in the USA are on average μ = 5 feet 11 inches tall, with a standard deviation of σ = 3 inches. Consider a sample of 1000 men from this population. Assuming a normal distribution, we predict that 623 (62.26%) will be between 5 ft. 8 in. and 6 ft. 2 in. ().

Where did the 5 ft. 8 in. and the 6 ft. 2 in. come from /? We add or subtract multiples of standard deviations. Thus, 6 ft. 2 in. = 5 ft. 11 in. (replace σ = 3) and 5 ft. 8 in. = 5 ft. 11 in. (again, replace σ = 3″).

Can generalize to any distribution. Chebyshev’s inequality (or theorem) guarantees that no more than a particular fraction of observations can be a specified standard deviations distance away from the mean ( needs to be greater than 1). Thus, for standard deviations, we expect a minimum of 75% of values within () two standard deviations away from the mean, or for , then 89% of values will be within () three standard deviations from the mean.

Hopefully you are now getting a sense how this knowledge allows you to plan an experiment.

For example, for a sample of 1000 observations of height of men, how many do we expect to be greater than 6 ft. 7 in. tall? Apply the empirical rule and do a little math — 79 inches (6 ft. 7 in) minus 71 inches (the mean) is equal to 8. Divide 8 by 3 (our value of ) you’ll get 2.66667; so, 6 ft. 7 in. tall is about 2.67 X standard deviations greater than the mean. From Chebyshev’s inequality we have , or 14% of observations less than or greater than the mean ().

Our question asks only about expected number of observations greater than ; divide 14% in half — we therefore expect about 70 individuals out of 1000 men sampled to be 6 ft. 7 in. or taller. Note we made no assumption about the shape of the distribution. If we assume the sample of observations comes from the normal distribution, then we can apply the Z score: about 4 individuals out of 1000 expected to be from the mean (see R code in Note box). We extend the Z-score work in Chapter 6.7.

round(1000*(pnorm(c(79), mean=71, sd=3,lower.tail=FALSE)),0)

round(1000*(pnorm(c(2.67), mean=0, sd=1,lower.tail=FALSE)),0)

[1] 4

Another example, this time with a data set available in R, the diabetic retinopathy data set in the survival package.

data(diabetic, package="survival")

Ages of the 394 subjects ranged from 1 to 58. Mean was 20.8 + 14.81 years. How many out of 100 subjects do we expect greater than 50 years old? With Chebyshev’s inequality we have , or 25.7% of observations less than or greater than the mean, so about 13. If we assume normality, then Z score (Upper tail) is 28.5% and we expect 28 subjects older than 50. Checking the data set, only eight subjects were older than 50. Our estimate by Chebyshev’s inequality was closer to the truth. Why /? Take a look at the histogram (Ch4) of the ages (Fig 3).

Figure 3. Histogram of ages of subjects in the diabetic retinopathy data set in the survival package.

Doesn’t look like a normal distribution, does it/?. 

Z-score or Chebyshev’s inequality, which to use /? Chebyshev’s inequality is more general — it can be used whether the variables is discrete or continuous and without assumption about the distribution. In contrast, the Z score assumes more is known about the variable: random, continuous, drawn from a normally distributed population. Thus, as long as these assumptions hold, the Z score approach will give a better answer. Makes intuitive sense — if we know more, our predictions should be better.

In summary from the above points, and perhaps more formally for our purposes, the standard deviation is a good statistic to describe variability of observations on subjects, it’s integral to the concept of precision of an estimate and is part of any equation for calculating confidence intervals (CI). For any estimated statistic, a good rule of thumb is to always include a confidence interval calculation. We introduced these intervals in our discussion of risk analysis (approximate CI of NNT), and we will return to confidence intervals more formally when we introduce t-tests.

When you hear people talk about “margin of error” in a survey, typically they are referring to the standard deviation — to be more precise, they are referring to a calculation that includes the standard deviation, the standard error and an accounting for confidence in the estimate (see also Chapter 3.5 – Statistics of error). 

Corrected Sums of Squares.

Now, take a closer look at the sample variance formula. We see a deviate

The variance is the average of the squared deviations from the mean. Because of the “squared” part, this statistic will always be positive and greater (or typically, equal to) zero. The variance has squared units (eg, if the units were grams, then the variance units are grams²). The sample standard deviation has the same units as the quantity (ie, no “squared” term). The numerator is called a sums of squares, and will be abbreviated as SS. Much like the deviate, will show up frequently as we move forward in statistics (eg, it’s key to understanding ANOVA).

Other standard deviations of means.

Just as we discussed for the arithmetic average, there will be corresponding standard deviations for the other kinds of means. With the geometric mean, one would calculate the standard deviation of the geometric mean, s_gm, as

where exp is the exponential function, ln refers to natural logarithm, and refers to the sample geometric mean.

For the sample harmonic mean, it turns out there isn’t a straight-forward formula, only an approximation (which I will spare you — it involves use of expectations and moments).

Base R, and therefore Rcmdr, doesn’t have built in functions for these, although you could download and install some R packages which do (eg, package NCStats, and the function is geosd() ).

If we run into data types appropriate for the geometric or harmonic means and standard deviations I will point these out; for now, I present these for completeness only.

Coefficient of variation (CV).

An unfortunate property of the standard deviation is that it is related (= “correlated”) to the mean — a fact recognized by Pearson (1896, p 276). Thus, if the mean of a sample of 100 individuals is 5 and variability is about 20%, then the standard deviation is about 1; compare to a sample of 100 individuals with mean = 25 and 20% variability, the standard deviation is about 5. For means ranging from 1 to 100, here’s a plot to show you what I am talking about.

Figure 4. Scatter plot of the standard deviation (StDev) by the mean. Data sets were simulated.

The data presented in the Figure 4 graph were simulated. Is this bias, a correlation between mean and standard deviation, something you’d see in “real” data? Here’s a plot of height in inches at withers for dog breeds (Fig 5). A line is drawn (ordinary linear regression, see Chapter 17) and as you can see, as the mean increases, the variability as indicated by the standard deviation also increases.

Figure 5. Plot of the standard deviation by the mean for heights of different breeds of dogs.

So, to compare variability estimates among groups when the means also vary, we need a new statistic, the coefficient of variation, abbreviated CV (Pearson 1896). The coefficient of variation is also known as the relative standard deviation, RSD. Many statistics textbooks not aimed at biologists do not provide this estimator (eg, check your statistics book).

This is simply the standard deviation of the sample divided by the sample mean, then multiplied by 100 to give a percent. The CV is useful when comparing the distributions of different types of data with different means.

Many times, the standard deviation of a distribution will be associated with the mean of the data.

Example. The standard deviation in height of elephants will be larger on the centimeter scale than the standard deviation of the height of mice. However, the amount of variability RELATIVE to the mean may be similar between elephants and mice. The CV might indicate that the relative variability of the two organisms is the same.

The standard deviation will also be influenced by the scale of measurement. If you measure on the mm scale versus the meter scale the magnitude of the SD will change. However, the CV will be the same!

By dividing the standard deviation by the mean you remove the measurement scale dependence of the standard deviation and generally, you also remove the relationship of the standard deviation with the mean. Therefore, CV is useful when comparing the variability of the data from distributions with different means.

One disadvantage is that the CV is only useful for ratio scale data (ie, those with a true zero).

The coefficient of variation is also one of the statistics useful for describing the precision of a measurement. See Chapter 3.4 Estimating parameters.

Standard error of the mean (SEM).

All estimates should be accompanied by a statistic that describes the accuracy of the measure. Combined with the confidence interval (Chapter 3.4), one such statistic is called the standard error of the mean, or SEM for short. For the error associated with calculation of our sample mean, it is defined as the sample standard deviation divided by the square root of n, the sample size.

sem →

The concept of error for estimates is a crucial concept. All estimates are made with error and for many statistics, a calculation is available to estimate the error (see Chapter 3.4).

Although related to each other, the concepts of sample standard deviation and sample standard error have distinct interpretations in statistics. The standard deviation quantifies the amount of variation of observations from the mean, while the standard error quantifies the difference between the sample mean and the population mean. The standard error will always be smaller than the standard deviation and is best left for reporting accuracy of a measure and statistical inference rather than description.

Questions.

1. For a sample data set like myY <- c(1,1,3,6), you should now be able to calculate, by hand, the
   • range
   • mean
   • median
   • mode
   • standard deviation
   • coefficient of variation
2. For our example data set, myX <- c(4,4,4,4,5,6,6,6,7,7,8,8,8,8,8) calculate
   • IQR
   • sample standard deviation, s
   • coefficient of variation
3. Use the sample() command in R to draw samples of size 4, 8, and 12 from your example data set stored in myX. Repeat the calculations from question 4. For example x4 <- sample(x,4) will randomly select four observations from x, and will store it in the object x4, like so (your numbers probably will differ!) x4 <- sample(x,4); x4 [1] 8 6 8 6
4. Repeat the exercise in question 4 again using different samples of 4, 8, and 12. For example, when I repeat sample(x,4) a second time I get sample(x,4) [1] 8 4 8 6
5. For Table 1, determine how many multiples of the standard deviation for observations greater than 95-percentile (eg, determine the observation value for a person who is in the 95-percentile for Height in the different decades, etc.
6. Calculate the sample range, IQR, sample standard deviation, and coefficient of variation for the following data sets
   • Height in inches of mothers,
   mom <- c(67, 66.5, 64, 58.5, 68, 66.5) #(data from GaltonFamilies in R package HistData)
   and fathers,
   dad <- c(78.5, 75.5, 75, 75, 74, 74) #(data from GaltonFamilies in R package HistData)
   • Carbon dioxide (CO₂) readings from Mauna Loa for the month of December for
   years <-c (1960, 1965, 1970, 1975, 1980, 1985, 1990, 1995, 2000, 2005, 2010, 2015, 2020)
   years <-c (1960, 1965, 1970, 1975, 1980, 1985, 1990, 1995, 2000, 2005, 2010, 2015, 2020) #obviously, do not calculate statistics on years; you can use to make a plot
co2 <- c(316.19, 319.42, 325.13, 330.62, 338.29, 346.12, 354.41, 360.82, 396.83, 380.31, 389.99, 402.06, 414.26) #(data from Dr. Pieter Tans, NOAA/GML (gml.noaa.gov/ccgg/trends/) and Dr. Ralph Keeling, Scripps Institution of Oceanography (scrippsco2.ucsd.edu/))
   • Body mass of Rhinella marina (formerly Bufo marinus) (see Fig. 1),
   bufo <- c(71.3, 71.4, 74.1, 85.4, 85.4, 86.6, 97.4, 99.6, 107, 115.7, 135.7, 156.2) 

Quiz – Chapter 3.3

Measures of dispersion

Start Quiz

Question

Confirm

✕

✓

Your answer:

Correct answer:

Next

Please wait..

↺

You got {{SCORE_CORRECT}} out of {{SCORE_TOTAL}}

Your Answers

Chapter 3 contents

• • Exploring data
  • Data types
  • Measures of Central Tendency
  • Measures of dispersion
  • Estimating parameters
  • Statistics of error
  • References and suggested readings

2.2 – Why do we use R Software?

If this page is TL;DR, and you’re in a rush, then instructions to install R onto your macOS or Windows PC are provided in Install R.  To run in the cloud only, see Use R in the cloud.

Why R? The case for using R in biostatistics education.

.

Why do we use R Software? Or put another way: DrD, Why are you making me use R?

Truth? You can use just about any acceptable statistical application to get the work done and achieve the learning objectives we have for beginning biostatistics. However, we will use the R statistical language as our primary statistical software in this course. Part of the justification is that all statistical software applications come with a learning curve, so you’d start at zero regardless of which application I used for the course.

In selecting software for statistics I have several criteria. The software should be:

• if not exactly easy, the software should have a reasonable learning curve
• widely accessible and compatible with all most personal computers
• well-respected and widely used by professionals
• free software
• open source
• well-supported for the purposes of data analysis and data processing
• really good for making graphics, from the basics to advanced
• capable to handle diverse kinds of statistical tests

R meets all of these criteria. R history began back in 1993 and has always been available as free software under the terms of the Free Software Foundation’s GNU General Public License in source code form. R compiles and runs on a wide variety of UNIX platforms and similar systems, including GNU/LINUX, FreeBSD, and various Linux distros like the popular Ubuntu^®, in addition to their more famous Microsoft Windows^® and Apple macOS^® distributions. To facilitate access to the software, numerous mirror sites are available from sites around the world, with cloud.r-project.org supported by RStudio perhaps the most widely used. From January 2024 to December 2024, more than 8.5 million downloads of base R were made from the RStudio CRAN mirror site (CRAN stands for Comprehensive R Archive Network.

install.packages("cranlogs")
library(cranlogs)
# How many downloads of base R around start of Fall semester?
 out <- cran_downloads("R", from = "2025-08-01", to = "2025-08-31")
sum(out$count)

[1] 582524

R is straightforward to use once you learn how to work with the language, but has a steep learning curve; after all, it’s a programming language. I recommend installing the program onto your computer, but ready to go R is also available at multiple cloud services, including Google CoLab and Posit Cloud. The remainder of this page is directed at users who choose to install R to their computer, although much of the discussion holds regardless of the R environment choice, local or Cloud.

For local installs of R, I also recommend use of the GUI R Commander.  Use of R Commander helps with the learning curve, and eventually, your use of code will become second nature. We don’t use RStudio, but as your skills improve, I recommend transitioning to the IDE RStudio; after the initial growing pains are behind you, RStudio likely will be a better solution over R Commander.

This discussion implies learning R programming. However, while we need statistical software to do statistics, students in my BI311 course must keep in mind that learning objectives for a biostatistics course is about the concepts and interpretation of statistics, not just use of the software. In other words, learning how to use R is not the focus of BI311 nor will you likely achieve R programming competency by the end of the semester. I certainly encourage students to strive for competency and I give frequent bonus opportunities to demonstrate coding skills during the semester.

Why R and not some other statistics software?

Thus, you might ask if the purpose of the course isn’t to learn R, why work with R instead of a more familiar app or software, eg, Microsoft Excel^® (hereafter simply referred to as Excel), or Google Sheets, or even my favorite open-source office alternative, LibreOffice Calc? Or, perhaps even just one of the many online calculators, if the course learning objective is to “just” learn about statistics?

First, I believe that real data derived from real biology or biomedical problems are essential elements to a first course in biostatistics. That’s not a particularly unique perspective although I don’t have survey results of other statistics instructors to back up the claim. Real problems involve observations on multiple subjects, many variables — large data sets; this alone precludes use of hand calculations and calculators. As a corollary, we will not spend a great deal of time learning the in’s and out’s of the algorithms that form particular statistical tests. Now, do understand that there is a tremendous benefit to understanding statistics by working through the equations, by looking at the algorithms, and there’s no escaping the need for understanding that probability provides the foundation of statistics inference (Chapter 8). Thus, for most of us, the statistical software available to us provides an appropriate framework for applying correct statistical tests to our projects. Therefore, the decision is about which statistical package we should use.

Second, R is perhaps the choice in academia for statistical software. A PUBMED search found more than 1500 citations of R. Visit Robert A. Muenchen’s web page (The popularity of data analysis software, r4stats.com) to see updated statistics on statistical software use. Those of you continuing on to graduate school or to professional schools will find that many of your statistically literate colleagues use R and not one of the commercial programs. While there are many excellent commercial packages (Table 1), and in some cases you can make spreadsheet programs do statistics (typically add-ins are required), all statistical software come with steep learning curves. Thus, part of my selling point to you is that learning to use R is at the cutting-edge in your field and, given that all of the software you could use can have have their challenges, it is best to work with something that will be around and is in wide use, without the burden of a financial investment.

Table 1. A selective list of statistical software.

see Wikipedia for list of additional software

Third, what about online sites like plot.ly where, for free, you can plot and, in some cases, calculate statistics? What about the web application at Brightstat, which claims to provide an SPSS-like experience online (Stricker 2008)? While it is true that there are many wonderful websites that can perform many of the statistical tests we will use this semester, these sites are not suitable for more than occasional use.

How to get started with R.

.

A few words about my approach to the software in my biostatistics course. I note to students during the semester that our course is not a programming course, but rather, programming skills are acquired along the way. Our focus is how to do statistical analysis, how to design experiments, etc. Coding skills without understanding of the why we’re doing it is not unlike learning to cook by following a great recipe as opposed to learning about the art and science of flavor (NPR Science Friday, March 15, 2024) — the homework is solved but the why we can support our conclusions may be obscure.

We set aside the first couple of class meetings simply to setup student’s Windows PC or Apple macOS computers to do data science; for students with access to Chromebooks or iPad tablets, we need to set them up to run R in the Cloud (eg., Google Colaboratory) — in other words, there’s a lot of pre-work that can present barriers to actual doing statistics in R (or other software). Once we have the environment set up for the students, we utilize a “tell-show-do” approach to learning how to code. The “tell” part includes sharing script; the “do” part is mostly done by students outside of class as part of completing homework.

First work with the language can be frustrating, so I also encourage a “five minute” rule — if it ain’t working, stop. So, by all means when Run through a troubleshooting checklist, but please don’t continue to struggle — the gap is between the instructions and where the student is with the material — it’s a me-as-instructor problem, not a student must work harder problem! 

All statistics work follows a similar workflow.

The basic workflow for a data project, using R as the primary statistical programming tool, looks like that illustrated in Figure 1.

Figure 1. A basic workflow with R.

The workflow begins with the assumption that a data set already exists. Getting the data into an R session depends on the source of the data, for example stored in spreadsheet file or from a table on a webpage. Data cleaning and processing of the data may be in order — for small and modest-sized data sets, my own preference is to do most of the data cleaning in my spreadsheet app, but R has full capabilities to handle or “wrangle” data. Next, the raw data should be stored in R session in a data frame object. Once the data frame object is available, exploratory data analysis, including summary statistics and data visualization can proceed. Higher order analytics, including statistical inference and modeling, can then begin.

The R statistical language, accompanied by additional packages to extend its capabilities beyond basic math and statistical functions, provided a complete statistical environment. R is best viewed as a programming language for statistics (data analysis), and data processing. Power users of R learn how to write scripts that do t-tests, ANOVA, regression, etc. The scripts are just lines of code that R understands and it provides the user tremendous control over analysis and inference of data sets. Because of this flexibility and power, however, R can be intimidating at first. So, we’ll start slowly with scripts, introducing just what we need to get started and build from there. We’ll be addressing R issues in more depth over the next several weeks, but for the first week, our goal(s) should be to make sure each of you knows how to start/exit R, how to create and utilize a working directory, and how to use R as a calculator. You obtain your copy of R from the R Project for Statistical Computing, available at https://www.r-project.org. Instructions to install R are provided in Install R.  A ten-part tutorial to get started using R is provided in Mike’s Workbook for Biostatistics.

 /Users/username/Desktop/BI311

Why R Commander, why not RStudio?

.

We utilize an R package that provides a menu-driven context to much of the typical statistics one needs to do biostatistics. The package is called R Commander (Rcmdr), which provides a graphic user interface or GUI. Rcmdr therefore significantly eases the learning curve for doing statistics with R. We use a package called R Commander, which provides drop down menus for most of the typical kinds of analyses. Rcmdr is in use in many courses across the world (more than 17K downloads in August 2025), and among the other GUI available for R, Rcmdr is among the best supported GUI available for R. R Commander function is extended by plug-ins; as of August 2025, there were 31 plugins that extend Rcmdr’s capabilities. Instructions to install R Commander are provided in Install R Commander. 

Wait! Why don’t we use Excel? My instructor in {insert course here} used Excel…

.

A very reasonable question for you to ask — why don’t we use Microsoft Excel or Google Sheets for statistics? Moreover, it is highly likely that you have gained at least some introduction to descriptive statistics and graphing with spreadsheets in former courses — shouldn’t we learn statistics within a software framework you are already familiar?

After all, “Can’t Microsoft Excel do statistics?” Mostly the answer is, no, not really (Fig 2).

Figure 2. “Spreadsheets,” xkcd.com no. 2180.

MS Excel, Google Sheets, Apple Numbers, and for that matter, Calc, the spreadsheet application in my favorite office app LibreOffice (LibreOffice is a free, open-source alternative to Microsoft Office), can be used to calculate many descriptive statistics. With some effort, these applications can be extended by use of either Analysis ToolPak or Solver Add-ins to do more complicated statistics like regression and analysis of variance, and curve fitting.

However, use of MS Excel for statistical analysis involves learning a number of commands, syntax, and developing workflows that are neither intuitive nor standard. Some publishers have provided add-ins that are reportedly designed to simplify this process (eg, XLStat, UNISTAT, Real Statistics using Excel). None of these options are free and none are in use in any major way by scientists (see Robert Muenchen’s The popularity of data analysis software). The free add-ins of Analysis ToolPak and Solver may work for you if you own a Windows PC, but only Solver is included for the Mac versions of Excel. MacOS users may download and install StatPlus:MacLE, which is a limited, but free alternative to the Analysis ToolPak add-in; for a complete package a Pro version is available (licenses started at $89, web site: www.analystsoft.com/en/products/statplusmacle/).

An additional caution: you should be aware that there have been reports over the years that algorithms selected by Microsoft for Excel have not always been to industry standards (eg, McCullogh and Wilson 2005). In short, the fit of Excel and other spreadsheet apps for use in statistics is not a simple one. To do the kinds of statistics we will use routinely in class, Excel would need to be modified with add-ins, and the add-ins would be the result of programming by someone. And you would still need to learn how to write the code.

What about graphics? You may like Microsoft Excel’s ability to do graphics. Indeed, Excel, Google Sheets, and LibreOffice Calc can be used to generate many typical kinds of statistical plots. But again, in comparison to R, spreadsheet app graphics are limited and require a deal of effort to generate acceptable plots. I think you’ll be surprised at how straight-forward R is. Here’s an example, first rendered in Microsoft Excel, then in base R. And importantly, the kinds of plots Excel does well at are not necessarily the plots suitable for research publication. For example, Excel allows you to make bar charts easily, but cannot do box plots. Box plots are preferred over bar (column) charts for ratio scale data.

Statistics comparisons between R and MS Excel.

.

About that learning curve. Let’s compare R and MS Excel for basic functions common in data analysis. Similar conclusions hold for comparisons to Google Sheets and LibreOffice Calcs.

Table 2 lists the observations we can use to conduct comparisons of the applications.

A simple data set

One of the first steps in data analysis is to produce what are called descriptive statistics. Common descriptive statistics are the mean and the sample standard deviation. Let’s compare Excel and R for retrieving these two statistics.

With Excel, to calculate the arithmetic mean of 24 numbers, enter the values into a single column of 24 rows, then enter “=average(A2:A25)“, without the quotes, into a new cell of the spreadsheet. “A2:A25” refers to where data would be contained in column A rows 2 through 25. Typically the first row in a worksheet would contain the name of the variable, eg, “A.” Depending on the significant figures set, the estimate returned by Excel for the mean of A is 59.58333333.

Similarly, to obtain the standard deviation, type =stdev(A2:A25), into a new cell of the spreadsheet. Again, depending on the significant figures set, Excel returns a value of 37.05215674 for the standard deviation of A.

In contrast, to obtain the mean and standard deviation for a variable in an R data set, all you would type at the R prompt (>), or in the script window

and then submit the code is:

myA <- c(12, 14, 20, 25, 28, 29, 32, 34, 35, 39, 47, 47, 50, 53, 54, 71, 79, 87, 89, 96, 105, 122, 130, 132)

where the “c” is a function to combine arguments into a vector and saved to the object myA, followed at the new line by

mean(myA)

Hit enter after entering the command) and R returns

[1] 59.58333

For the standard deviation, write the R base function sd()

sd(myA)

Hit enter after entering the command and R returns

[1] 37.05216

It’s not much of a difference, but note that to get the mean (arithmetic average) I typed seven characters in R, but 16 characters in Excel; similarly, for the standard deviation I typed in 5 characters in R, but 13 characters in Excel. That’s a savings of 56% and 62%, respectively. Excel tries to help by using AutoComplete to anticipate what you want to enter, but AutoComplete doesn’t always work properly (eg, see gene name errors generated by use of default Microsoft Excel settings, Ziemann et al 2016).

In conclusion, R is quicker for descriptive statistics.

Graphics comparison between R and MS Excel.

.

MS Excel is often cited for its graphics capabilities (Camões 2016). We can make the familiar scatter plots, bar charts, and pie charts in Excel. These plots and more are easily obtained in R. I won’t elaborate here about graphics, we talk at some length about graphics in Chapter 4. But here’s one example in R.

Let’s plot myB vs myA. We already provided the data for variable A, here’s the data for variable B.

17, 21, 21, 26, 27, 32, 28, 42, 40, 30, 71, 53, 56, 61, 55, 89, 82, 63, 116, 162, 116, 154, 137, 149

Don’t recall how to assign a set of numbers to an object, B, in R? See above and look again at how we assigned the numbers to object myA.

To get a simple scatter plot (Fig 3), I may write at the R prompt.

plot(myA,myB)

Figure 3. Basic scatter plot made in R, plot(myA,myB).

And here’s the comparable default plot (Fig 4) from Microsoft Excel, Office 365

Figure 4. Basic scatterplot made in Microsoft Excel.

Now, both graphs need some work, and to be fair, these are just the defaults. With some effort, you can make an Excel graph look pretty good. But note — the defaults in Excel don’t generate axis labels, while R default plot does. Excel adds a useless title and legend; both need to be removed. Excel also adds grid lines where typically one would not include these in a scientific plot; for example, in the case of scatterplots, grid lines can distract the viewer from the data points and reduce the data-to-ink ratio Tufte 1984).

Count the steps to generate an acceptable scatter plot (Table 3). I’ve also added R Commander (Rcmdr) steps for comparisons (Rcmdr lets you use drop-down menus like Excel or Google Sheets or LibreOffice Calcs).

Table 3. Steps needed to make a simple scatterplot in R, R Commander, or Microsoft Excel.

Conclusion? R is quicker for routine statistical plots like a scatter plot. And I didn’t even count the steps needed to change MS Excel’s dreadful diamond icon points.

That’s one step in R, four steps in Rcmdr, but eight steps for Microsoft Excel. LibreOffice Calc is a little better at only four steps, but like MS Excel, you’d need to change several components to the graph (Fig 5).

Figure 5. Basic scatterplot made in LibreOffice Calc.

plot(myA,myB, pch = 22)

So, you’re telling me I don’t need a spreadsheet application?

.

No, not at all. We use spreadsheets, and more generally, databases, to store data. Spreadsheets apps are designed to make data entry and data management approachable and efficient. They remain an important tool for researchers (Browman and Woo 2017).

R is not that great of a spreadsheet; packages are available to seamlessly tie your spreadsheet and database data to R via ODBC. We will routinely enter and manipulate data in MS Excel, then import the data into R for analysis.

Spreadsheet apps like MS Excel and Google Sheets (see also LibreOffice Calc) are great at being a spreadsheet program, R is great at being a statistical software program. You should take advantage of what the tools do best.

Still not convinced?

.

R is in use all over the world, by students and professionals alike and if one is going to spend the time to learn how to use a statistics software program, you should learn a standard program, like R.

And it’s not just me. Read about R in this 2009 New York Times piece, “Data analysts captivated by R’s power.” Look who purchased (April 2015) Revolution Analytics, a major player in the development of the R programming language.

Why install R on your computer?

.

Convenience. Control. Offline.

At the Biology department of Chaminade University, we have installed and maintain R, Rcmdr, and RStudio along with all required packages on our Macbook Pro® Lab computers for your use during class and during optional, proctored biostatistics work sessions. Since 2018, R is increasingly available “in the cloud” (eg, RStudio Cloud), which would mean you could run R in your browser and avoid installation on your computer. You can run significant analysis with R in the cloud via the free Google Colaboratory and CoCalc are now available: I encourage you to look into these platforms. Unfortunately, these services are not quite ready for the classroom. For example, RStudio in the Cloud is free to use on a limited basis, but quickly requires a significant subscription cost with increasing use. Google Colab and GoCalcs require use of Jupyter notebooks, which add yet another layer to the learning curve without focusing on learning statistics. Second, although access to their servers is easy, running simultaneous connections via Chaminade’s single public IP address is likely to lead to problems for us. Third, I want you to use R Commander (Rcmdr) to assist in the learning curve — Rcmdr cannot be run in the Cloud (ie, RStudio in the Cloud, Google Colaboratory, or CoCalc).

Therefore, you are encouraged to install R, Rcmdr, and even RStudio, onto your own computers, in part because of the convenience, but also because R is not generally available to students on campus, ie, only the Biology department’s computers have the up-to-date R software installed.

To get started, go to your Canvas website and view How to install R on your own computer.

An additional benefit to installing a version of R on your computer, you’ll understand more about the software if you take the time to install and if need be, troubleshoot your installation of the software. Moreover, there’s a considerable amount of help out there for R. For example, a simple Google search(keywords: tutorial “install R”), returns more than 700K hits, and more than 40K January 2023 alone (add “after:2023-01-01” to Google search box). In fact, there’s so much out there that you’ll want to sample from several sites and select the voice that works best for you.  

Questions

1. Write up three learning outcomes for this page. Hint: Point your favorite generative AI to this page and ask for help.
2. I listed several spreadsheet apps: Microsoft Excel, Google Sheets, Apple Numbers, LibreOffice Calc. Which of these are free to download and install to your computer? Which are freely available via the Cloud?
3. What level of confidence do you have to this statement: I am confident in my ability to use spreadsheet apps for my BI-311 work this semester. Response options: 
   1. Strongly disagree.
   2. Disagree.
   3. Agree.
   4. Strongly agree.
4. What level of confidence do you have to this statement: I am confident in my ability to use the R programming software for my BI-311 work this semester. Response options: 
   1. Strongly disagree.
   2. Disagree.
   3. Agree.
   4. Strongly agree.

.

Quiz Chapter 2.2

Why do we use R software?

Start Quiz

Question

✕

✓

Your answer:

Correct answer:

Next

Please wait..

↺

You got {{SCORE_CORRECT}} out of {{SCORE_TOTAL}}

Your Answers

Chapter 2 contents

.

• Introduction
• Why biostatistics?
• Why do we use R Software?
• A brief history of (bio)statistics
• Experimental Design and rise of statistics in medical research
• Scientific method and where statistics fits
• Statistical reasoning
• Chapter 2 – References and suggested reading

4 – How to report statistics

Introduction: Reporting your results.

While you are thinking about exploring data sets and descriptive statistics, please review our overview of data analysis (Chapter 2.4 and 2.5). While the scientific hypotheses come first, how experiments are designed should allow for straight-forward analysis: in other words, statistics can’t rescue poorly designed experiments, nor can it reveal new insight after the fact.

Once the experiments are completed, all projects will go through a similar process:

• Description: Describe and summarize the results.
• Check assumptions.
• Inference: conduct tests of hypotheses.
• Develop and evaluate statistical models.

Clearly this is a simplification, but there’s an expectation your readers will have about a project. Basic questions like how many subjects got better on the treatment? Is there an association between Body Mass Index (BMI) and the primary outcome? Did male and female subjects differ for response to the treatment? Undoubtedly these and related questions form the essence of the inferences, but providing graphs to show patterns may be as important to a reader as any p-value — a number which describes how likely it is that your data would have occurred by chance — eg, from an Analysis of variance.

Each project is unique — sorry, what a breathtakingly obvious statement — but what elements must be included in a results section?

Data visualization.

We describe data in three ways: graphs, tables, and in sentences. In this page we present the basics of when to choose a graph over presenting data in a table or as a series of sentences (ie, text). In the rest of this chapter we introduce the various graphics we will encounter in the course. Chapter 4 covers eight different graphics, but is by no means an exhaustive list of kinds of graphs. Phylogenetic network graphs are presented in Chapter 20.11. Although an important element of presentation in journal articles, we don’t discuss figure legends or table titles; guidelines are typically available by the journal of choice (eg, PLOS ONE journals guidelines).

A quick note about terminology. Data visualization encompasses charts, graphs and plots. Of the three terms, chart is the more generic. Graphs are used to display a function or mapping between two variables; plots are kinds of graphs for a finite set of points. There is a difference among the terms, but I confess, I won’t be consistent. Instead, I will refer to each type of data visualization by its descriptive name: bar chart, box plots, histograms, scatter plot, etc. Note that technically, a scatter plot can refer to a graph, eg, a line drawn to reflect a linear association between the two variables, whereas bar charts and box plots would not be a graph because no function is implied.

Why display data?

Do we just to show a graph to break the monotony of page after page of text or do we attempt to do more with graphs? After all, isn’t “a picture’s worth a thousand words?” In many cases, yes! Graphics allow us to see patterns. Data visualization is a key part of exploratory data analysis (Tukey 1977), or data mining in the parlance of big data (Keim 2001). Data visualization is also a crucial tool in the public health arena, and finding effective graphics to communicate, at times, complex data and information to both public and professional audiences can be challenging (see discussion in Meloncon and Warner 2017). 

Graphics are complicated — all the interplay among pixels etc and rendering process — and expensive to do well. Text is much cheaper to publish, even in digital form. But the ability to visualize concepts, that is, to connect ideas to data through our eyes (see Wikipedia), seems to be more the cognitive goal of graphics. Lofty purpose, desirable goal. Yes, it is true that graphics can communicate concepts to the reader, but with some caution. Images distort, and default options in graphics programs are seldom acceptable for conveying messages without bias (Glazer 2011). 

Here’s some tips from a book on graphical display (Tufte 1983; see also Camões 2016).

Your goal is to communicate complex ideas with clarity, precision, and efficiency. Graphical displays should:

• show the data
• avoid distorting the data
• present numbers in a small space
• help the viewer’s eye to compare different pieces of data
• serve a clear purpose (description, exploration, tabulation, decoration)
• be closely integrated with statistical and verbal descriptions of a data set.

We accomplish these tasks by following general principles involving scale and a commitment to avoiding bias in our presentation.

Importantly, graphs can show patterns not immediately evident in tables of numbers. See Table 1 for an example of a dataset, “Anscombe’s quartet,” (Anscombe 1973), where a picture is clearly helpful.

Anscombe’s quartet

The Anscombe dataset is also available in R packages data(Anscombe, package="carData"), or you can copy/paste from Table 1 into a spreadsheet or text file, then load the data file into R (eg, Rcmdr → Load data set). Note that the data set does not include the column summary statistics shown in the last row of the table.

Before proceeding, look again at the table — See any patterns in the table?

Maybe.… Need to be careful as we humans are really good at perceiving patterns, even when no pattern exists.

Now, look just at the last row in the table, the row containing the descriptive statistics (the means and standard deviations). Any patterns?

The means and standard deviations are the same, so nothing really jumps out at you — does that mean that there are no differences among the columns then?

But let’s see what the scatter plots look like before we conclude that the columns of Y ’s are the same (Fig. 1). I’ll also introduce the R package clipr, which is useful for working with your computer’s clipboard.

Figure 1. Scatter plot graphs of Anscombe’s quartet (Table 1)

#R code for Figure 1.

require(clipr)
#Copy from the Table and paste into spreadsheet (exclude last row). Highlight and copy data in spreadsheet
myTemp <- read_clip_tbl(read_clip(), header=TRUE, sep = "\t")
#Check that the data have been loaded correctly
head(myTemp)
#attach the data frame, so don't have to refer to variables as myTemp\$variable name
attach(myTemp)
#set the plot area for 4 graphs in 2X2 frame
par(mfrow=c(2,2))
plot(X, Y1, pch=19, col="red", cex=1.2)
plot(X, Y2, pch=19, col="orange",cex=1.2)
plot(X, Y3, pch=19, col="yellow",cex=1.2)
plot(X, Y4, pch=19, col="blue",cex=1.2)

And now we can see that the Y ‘s have different stories to tell. While the summary (descriptive) statistics are the same, the patterns of the association between Y values and the X variable are qualitatively different: Y1 is linear, but diffuse; Y2 is nonlinearly associated with X; Y3, like Y1, is linearly related to X, but one data point seems to be an outlier; and for Y4 we see a diffuse nonlinear trend and an outlier.

So, that’s the big picture here. In working with data, you must look at both ways to “see” data — you need to make graphs and you also need to calculate basic descriptive statistics.

And as to the reporting of these results, sometimes Tables are best (i.e., so others can try different statistical tests), but patterns can be quickly displayed with carefully designed graphs. Clearly, in this case, the graphs were very helpful to reveal trends in the data.

When to report numbers in a sentence? In a table? In a graph?

The choice depends on the message. Usually you want to make a comparison (or series of comparisons). If you are reporting one or two numbers in a comparison, a sentence is fine. “The two feral goat populations had similar mean numbers (120 vs. 125) of kids each breeding period.” If you have only a few comparisons to make, the text table is useful:

Table 2. Data from Kipahoehoe Natural Area Reserve, SW slope of Mauna Loa.

To conclude, tables are the best way to show exact numbers and tables are preferred over graphs when many comparisons need to be made. 

Couldn’t I use a pie chart for this?

Yes, but I will try and persuade you to not. Pie charts are used to show part-whole relationships. If there are just a few groups, and if we don’t care about precise comparisons, pie charts may be effective. Some good examples are Figure 1 O’Neill et al 2020, 

Sometimes, people use pie charts for very small data sets (comparing two populations, or three categories, for example). These work well, but as we increase the number of categories, the graphic likely requires additional labels and remarks to clarify the message. The problem with pie charts is that they require interpretation of the angles that define the wedges, so we can’t be very precise about that. Bar charts are much better than pie charts — but can also suffer when many categories are used (Chapter 4.1). 

To illustrate the problem, here’s a couple of pie charts from Microsoft Excel (a similar chart can be made with LibreOffice Calc) for our goat data set; compare this graph to the table and to the bar chart below (Fig 2).

Figure 2. Excel pie chart of Table 2 data set.

A bar chart of the same data (Fig 3).

Figure 3. Bar chart of Table 2 data set

The bar chart (Fig 3) is easier to get the message across; more goat kids were found outside the fenced area then inside the fenced in areas. We can also see that more goat kids were found in the “other” areas compared to the kipuka. The pie chart (Fig 2) in my opinion fails to communicate these simple comparisons, conclusions about patterns in the data that clearly would be the take-home message from this project. Aesthetically the bar chart could be improved — a mosaic plot would work well to show the associations in the project results (See Ch 4.4 Mosaic plots). Furthermore, a strong case can be made that the bar chart is over kill — after all, there’s only the four points to show.

But we are not done with this argument, to use graphics or text to report results. Neither the bar chart (mosaic plot) or the pie chart really work. The reader has to interpret the graphics by extrapolating to the axes to get the numbers. While it may be boring — 1.5 million hits Google search “data tables” boring — tables can be used for comparisons and make the patterns more clear and informative to the reader. Here’s a different version of the table to emphasize the influence of fencing on the goat population.

Table 3. Revised Table 2 to emphasize comparisons between inside and outside the fence line on feral goat population on Mauna Loa.

Table 3 would be my choice — over a sentence and over a graph. At a glance I can see that more goat kids were found outside of the fenced area, regardless of whether it was in a kipuka or some other area on the mountain side. Table 3 is an improvement over Table 2 because it presents the comparisons in a 2 X 2 format — especially useful when we have a conditional set.

For example, it’s useful to show the breakdown of voting results in tables (numbers of votes for different candidates by voters party affiliation, home district, sex, economic status, etc.). Interested readers can then scan through the table to identify the comparison they are most interested in. But often, a graph is the best choice to display information. One final point, by judiciously combining words, numbers, and images, you should be able to convey even the most complex information in a clear manner! We will not spend a lot of time on these issues, but you will want to pay some attention to these points as you work on your own projects.

Some final comments about how to present data.

What your graph looks like is up to you, lots of people have advice (eg, Klass 2012). But we all know poor graphs when we see them in talks or in papers; we know them when we struggle to make sense of the take-home message. We know them when we feel like we’re missing the take-home message.

Here’s my basic, very subjective take on communicating information with graphics — aka a grammar of graphics, sensu Tufte 2001, and Wilkinson 2005.

• Always include proper labeling: axes, units, legends, and captions.
• Minimize white space (for example, the scatter plots above could be improved simply by increasing the point size of the data points)
• Yes, minimize white space, but always respect the scale, particularly when comparisons are intended.
• Avoid bar charts for comparisons if you are trying to compare more than about three or four things.
• Use color sparingly and with intention: a consistent palette distinguishes groups.
• Avoid overly simple graphs with just a handful of points or bars. Just write it up!
• A graphic in a science report that is worth “a thousand words” probably is too complicated, too much information, and, very likely, whatever message you are trying to convey is better off in the text.

Clever graphics may be visually stunning and, perhaps, serve well to advertise the take home message, but likely they need to be accompanied by a sophisticated text as guide to those take home messages.

Homework to go with this topic.

Homework 2B: Graphs in Mike’s Workbook for biostatistics.

Quiz

In addition to ten general questions in quiz-form about this page contents, a total of 94 questions among the several subchapters, a mix of true or false and multiple choice question format. 

Quiz Chapter 4

How to report statistics

Start Quiz

Question

✕

✓

Your answer:

Correct answer:

Next

Please wait..

↺

You got {{SCORE_CORRECT}} out of {{SCORE_TOTAL}}

Your Answers

Chapter 4 contents

• Graphs and tables
• Bar (column) charts
• Histograms
• Box plot
• Mosaic plots
• Scatter plots
• Adding a second Y axis
• Q-Q plot
• Ternary plots
• Heat maps
• Graph software
• References and suggested readings