Analyzing Ratios of Prevalence and Incidence Estimators in Multilevel Mixed Effects Models

• Activity Number: 645
• Type: Contributed
• Date/Time: Thursday, August 7, 2014 : 10:30 AM to 12:20 PM
• Sponsor: Biometrics Section
• Author(s): William Johnson*+ and Jeff Burton and Robbie Beyl and Hongmei Han
• Companies: Pennington Biomedical Research Center and Pennington Biomedical Research Center and Pennington Biomedical Research Center and Pennington Biomedical Research Center
• Keywords: Binary data, Fixed and random effects models, Generalized linear mixed models, Log-linear regression

Abstract

Inferences pertaining to prevalence are of interest in cross-sectional population surveys whereas inferences concerning incidence typically are encountered in follow-up studies such as randomized control trials. Statistical analyses may be carried out in terms of various functions of prevalence and incidence estimators, but the focus here is, respectively, on prevalence ratio and relative risk (incidence ratio) estimators. Discussion is limited to data having a hierarchical structure such as that found in random cluster sampling and group randomized trials. For example, a sample of cities may be selected (Level 1) and within each selected city a sample schools may be selected (Level 2) and within each school a sample of students may be selected (Level 3). At any level the sampling units may be randomly allocated to intervention groups. Differential effects may be investigated for all levels and responses may be observed on cross-sectional or longitudinal outcomes. Mixed effects log-linear and logistic regression models are used to illustrate novel analytic approaches in illustrative pragmatic applications.