Analysis of Variance Method for Longitudinal Data

概要

論文の詳細を見る
Because a particular feature of longitudinal data is the repeated measurement of subjects over time, it inevitably produces correlation among observations. Therefore, it is not appropriate to use procedures developed under an assumption of independence. In the conduct of actual pharmaceutical clinical trials, due to problems of ethics, cost to the maker, etc. it is necessary to limit the scale of the pilot study by, for example, collecting repeated observations on the same subjects. Therefore, procedures for longitudinal data analysis are important in practice. In this paper, we consider a method of analysis for multivariate longitudinal measurements on multiple subjects in a fixed time period. We treat the individuals as random effects, as the purpose of the analysis is not to make inference about each individual per se but rather the population. For such a model, we give the analysis of variance and describe the parameter-estimation method to examine whether there are effects of group, phase and item.