|
Abstract:
|
Unbalanced repeated measures data can arise in many different areas of science, and many software packages now contain functions for model development and hypothesis testing for this type of data. Unfortunately, the methodology being used is often based on large samples and correctly specified models. These assumptions may be violated, creating the potential for a practitioner to draw questionable conclusions. Therefore, we propose a joint model of the mean and covariance for independently, but not identically distributed observation vectors. We consider a nonparametric mean function, and develop both bootstrap and asymptotic measures of uncertainty for unbalanced repeated measures. To illustrate our method, we present the results of a simulation study and data analysis that compare the bootstrap and asymptotic approaches in finite samples.
|
