M. Durban Reguera, G. Camarda
The Composite Link Model is an advanced framework used when each observation must be linked to a linear function of multiple predicted values. A key application of this methodology is estimating latent processes behind grouped data, treating them as indirect observations of underlying trends, highlighting the CLM’s ability to uncover the deeper structures within aggregated or incomplete data sets. This is common in demography and epidemiology, particularly in mortality estimation, where death counts are aggregated over ages, years, months or weeks. Here we will show how to estimate latent mortality patterns across time and age by imposing smoothness through a roughness penalty in the likelihood function. While effective, the model faces computational challenges in high-dimensional settings. To address this, we propose a reformulated iterative estimation procedure using Generalized Linear Array Models, enabling smooth disaggregation of latent distributions in multidimensional data.
Palabras clave: Grouped counts, Composite Link Model, Generalized Linear Array Models, Penalized splines, Mortality
Programado
Bioestadística I
13 de junio de 2025 09:00
MR 3