128 PART 3 Getting Down and Dirty with Data

Software can provide various enhancements to the basic box plot. Figure 9-8b

illustrates two such embellishments you may consider using:»

» ^{Variable width: The widths of the bars can be scaled to indicate the relative}

size of each group.»

» ^{Notches: The box can have notches that indicate the uncertainty in the}

estimation of the median. If two groups have non-overlapping notches, they

probably have significantly different medians.

Depicting the relationships between

numerical variables with other graphs

We started this chapter by developing summary statistics and making graphs of

one numeric variable at a time. One example was where we took seven measure-

ments of diastolic blood pressure (DBP) from a group of study participants and

developed summary statistics. This is called a univariate analysis because it only

concerns one variable. But in the example of box plots in the preceding section, we

conducted a bivariate analysis because we were looking at the relationship between

two variables in a sample of patients from four different clinics. The two variables

were enzyme levels, and source clinic (Clinic A, B, C, or D). We could have done

another bivariate analysis looking at two continuous variables (such as two differ-

ent enzyme levels in participants) using a scatter plot, which is covered thor-

oughly in Chapter 16.

This chapter focused on univariate and bivariate summary statistics and graphs

that can be developed to help you and others better understand your data. But

many research questions are actually answered using multivariate analysis, which

allows for the control of confounders. Being able to control for confounders is one

of the main reasons biostatisticians opt for regression analysis, which we describe

in Part 5 and Chapter 23. In these chapters, we cover the appropriate summary

statistics and graphical techniques for showing relationships between variables

when setting up multivariate regression models.