ISM Research Memorandum
No.
922
Title:
A Bayesian logit age-period-cohort model
Author(s):
Nakamura, Takashi (The Institute of Statistical Mathematics, Tokyo)
Key words:
ABIC; cohort analysis; identification problem; Japanese national character study; repeated social survey
Abstract:
Cohort analysis is a method of estimating the age, period, and cohort effects from a set of repeated social survey data. It has confronted, however, an identification problem in that the age, period, and cohort effects cannot be decomposed uniquely without some prior information. The author overcame the problem by introducing a Bayesian model with a gradually-changing-parameter assumption and a procedure to select the optimal model based on the Akaike Bayesian information criterion, ABIC. This paper presents the Bayesian logit age-period-cohort model and its application to the Japanese national character study data.