| Sriya Venkat |
SAS provides a wide range of procedures for modeling continuous, binary, and count data. In this paper, we will analyze
the uses of PROC REG, PROC GLM, PROC GENMOD, and PROC MIXED. These procedures, though they sometimes
have overlaps in output, significantly differ in their methods of estimation, complexity, and underlying assumptions. This
paper builds on previous comparative work and develops a practical guide for biostatisticians working with real-world
health data. Using examples from various clinical trials and epidemiological studies, we demonstrate how estimation,
interpretability, and fit are impacted by model choice - especially when we have causes of non-independence,
non-normality, and repeated measures. We evaluate these four procedures on case studies involving blood pressure
monitoring, adverse event counts, and longitudinal biomarker data. The result is a decision-support framework that helps
researchers select and apply the most appropriate procedure based on study design, outcome type, and expected
modeling outcomes. This paper aims to support training and transparency for analysts in the health sciences using SAS
for complex data structures. |