Back to Annual Meeting Page		American Public Health Association 133rd Annual Meeting & Exposition December 10-14, 2005 Philadelphia, PA

Abstract #117285

Chih-Ping Chou Chou, PhD, Lingyun Ji, and Bin Xie, MD. Department of Preventive Medicine, Keck School of Medicine, University of Southern California, 1000 S. Fremont Avenue, Unit 8, Alhambra, CA 91803, (626)457-6649, cchou@usc.edu

Age-Period-Cohort (APC) technique has been frequently utilized on cross-sequential data, in which same age groups have been observed longitudinally, to investigate longitudinal trend of phenomenon of interest by decomposing the potential impacts on secular trend into age, period of measurement, and/or birth cohort effects. Parameters specified in the APC model can be estimated using the SAS/GENMOD procedure. SAS/Macro has been developed to perform the estimation and to provide informative print outs for model comparisons. Two major issues encountered in the application of APC technique are model specification and model identification. The former refers to whether the appropriate model to reflect the correct sources of impacts has been selected, while the latter involves if appropriate reference points have been selected. A large literature has been devoted to the latter issue to solve the identifiability problem: the fact that there is a linear relationship among age, period and cohort. The purpose of this study is to investigate the sensitivity of the APC due to model misspecification and selection of reference points. Sensitivity analysis investigates factors that determine the fit of the model, and study the sensitivity of the fit to these factors. This study will focus especially on the changes of parameter estimates due to model misspecification and selection of reference points. Diagnostic plots will also be used for visual comparisons. Both simulated data and empirical data on smoking among youth will be used to evaluate this property of APC.

Learning Objectives: