Modeling Digit Preference by Penalized Composite Link Models
Carlo G. Camarda, Max Planck Institute for Demographic Research
Paul H. Eilers, Leids Universitair Medisch Centrum (LUMC)
Jutta Gampe, Max Planck Institute for Demographic Research
Age heaping is a typical result of digit preferences in demography. In this paper we present a method which overcomes this issue. The concept of penalized likelihood and the composite link model are combined. The only assumption for the age distribution is that counts for adjacent digits should be similar. The combined model allows extraction of both the latent distribution and the pattern of misreporting probabilities from and to specific ages. A simulation study is given to demonstrate the performance of the method. The age distribution of the population of the Philippines in 1960 is used as a typical example for an application to real data. We show that the combination of two statistical tools can provide supplementary information compared to the usual index measures.
Presented in Poster Session 1