Survival Analysis using Integrated Nested Laplace Approximations: A Case Study of House Sparrows from Aldra

Abstract

The animal model is a generalized linear mixed model (GLMM) often used in evolutionary biology and animal breeding. Integrated Nested Laplace Approximation (INLA) is a quite new methodology for making fast non-sampling based Bayesian inference for hierarchical Gaussian Markov models. In this report we show that INLA fit into a number of survival models. In particular, we consider a Weibull and a Semi-parametric model with time varying covariates. It is also shown that INLA is applicable for a logistic model and a version of the Bayesian animal model with binomial likelihood and logit link. Analysis of the animal model is done both with a simulation study and for real data coming from an inbred natural house sparrow population. From the simulation study we observe that INLA estimated the different model parameters with their 95% credible interval well, but it is worth noticing that the standard deviations are quite large. It is important to note that the work done in this report is driven by the nature of data available in the data set.