ベイズ型Age-Period-Cohort 分析の多項ロジットモデル --反復横断調査を用いた政治課題の要因分解--

The purpose of this study is to discuss the implementation and importance about the multinomial logit model of Bayesian Age-Period-Cohort analysis. Since the design matrix of three factors does not become full rank due to the identification problem derived from linear dependency between age, period and cohort, few studies have used a complex model. In contrast, this paper presents the practical multinomial logit model with efficient sampling by reparameterization about the constraints of assuming a random walk for each effect. Furthermore, we compared the estimates of the logit model and the multidimensional model about political issues that should be considered from multiple perspectives. Using repeated cross-sectional survey in Japan from 1973 to 2013, the multinomial logit model was shown to be effective because it was able to obtain a trend that were not captured in the logit model. [ABSTRACT FROM AUTHOR]