Analyzing longitudinal data by using population averaged and subject specific approaches

• Published in: AIP conference proceedings Vol. 3189; no. 1
• Main Authors: Hamzah, Nurfaten, Shaik Abdullah, Fauhatuz Zahroh
• Format: Journal Article Conference Proceeding
• Language: English
• Published: Melville American Institute of Physics 19.08.2024
• Subjects: Data analysis; Errors; Estimates; Inference; Parameter estimation; Statistical models
• ISSN: 0094-243X; 1551-7616
• DOI: 10.1063/5.0225236

Special techniques are required to analyze longitudinal data that includes the correlation of repeated measurements within each subject. In this study, the comparison of analyzing longitudinal data using the population-averaged approach and the subject-specific approach were done. Binary longitudinal dataset with five repeated measurements on 7827 subjects were analyzed using these two techniques. The study aims to model the association between a binary response variable, which is the emotional health problem of the elderly (depression), and a set of predictor variables which are gender, race, religious preference, age, perceived health, arthritis disease, educational status and marital status. Almost similar results were shown in terms of regression parameters estimates for the population-averaged approach with generalized estimating equations (GEE) and the subject-specific approach with generalized linear mixed models (GLMM). However, there were differences in the value of parameter estimates and standard errors. The regression parameters estimates and standard errors of population-averaged approach were smaller than subject-specific approach. Hence, due to their different interpretation, a population-averaged approach is appropriate when the target of inference is the population and a subject-specific approach is appropriate when the target of inference is the subject.