Data Evaluation and Statistical Analysis of Functional Observational Battery Data Using a Linear Models Approach

Abstract

Statistical analysis of functional observational battery (FOB) data presents special problems in that there are three different types of data collected (continuous, count, and categorical), all of which are measured in a repeated manner across time. Initial measurements are made before any treatment is applied, and proper use of these individual control values must be determined. A coherent structure for the analysis of such data is laid out, and examples of applications are presented. Rationale for the approaches used are described. Behavioral characteristics of the statistical tests are summarized for one FOB experiment to show that the tests indeed perform properly. The availability and ease of use of the SAS statistical software employed, including the key analysis procedures PROC CATMOD and PROC GLM, and of the SASGRAPH graphics procedures and their importance to data evaluation in the FOB are fully described. Cautions about these procedures and further statistical research and development needs are summarized.