ABSTRACT

In this chapter, the author focuses on two important issues of parameter estimation. One is to accurately calculate variances and standard errors based on the asymptotic results and establish the Delta method for the variance of the period and cohort effects in fitting a generalized linear model to the age-period-cohort (APC) data. The other is to select the side conditions by searching for efficient estimation with least variance through theoretical calculation and simulation studies. Although the parameter estimates of the intrinsic estimator can be efficiently computed using the principal component analysis (PCA) method for both the linear and loglinear APC models, the variance estimation from the PCA approach may not be valid for the loglinear models. The author provides the intrinsic estimator and the standard errors calculated by the PCA and Delta methods for the loglinear model fitted to the lung cancer mortality data among US males.