Abstract
SAS provides a macro GLIMMIX, which can be used for modelling of discrete spatial variation in epidemiological studies, where data are aggregated into small areas such as municipalities or postcode sectors. The purpose of these models is primary to examine to what extent unmeasured spatially correlated variables can explain the outcome of interest. Some necessary additional code is proposed for this macro implementing some of the most used models for analysing and exploring spatial variation, in for example Poisson and logistic regression: Gaussian intrinsic conditional autoregression and spatial multiple memberships models originated from multilevel models. The code is illustrated by analysing the well-known Scottish lip cancer dataset with GLIMMIX and the results are compared with a Markov chain Monte Carlo approach. The code gives epidemiologists and bio-statisticians an immediate tool for analysing discrete spatial models in a familiar statistical software package.
Published Version
Talk to us
Join us for a 30 min session where you can share your feedback and ask us any queries you have
Schedule a call