ABSTRACT
In this chapter, parametric techniques that model the mean of the repeated measures over time are developed, and there are many ways to estimate the mean profile for repeated measures, which were introduced in Chapter 4. First and foremost is computing the descriptive statistics (mean, median, mode, and standard deviation) of the response at each time point, which should be replicated for each covariate (e.g., treatment group and other values such as age). In addition to the descriptive statistics, a scatter plot with a lowess curve should be conducted, and the plot reconciled with the descriptive statistics, that is, the plot should corroborate the descriptive statistics. This will entail a give-and-take process, whereby some plots (e.g., varying the type of plot such as fitting linear, quadratic, and lowess curves) will agree more and some will agree less with the descriptive statistics. Thus, one’s choice of a parametric model (linear, quadratic, and linear shift point) is to a large extent subjective. That is, before one chooses a particular parametric model, one will have to determine the degree of agreement and various scatter plots of the data.
