Analysis of count data

Theme Co-ordinator: Neal Alexander

Overview

In medical research, some things we may be interested in counting are (with links to examples):

For example, we may want to identify factors are associated with higher disease rates or parasite density.  This may include spatial modelling.

The simplest model for count data is the Poisson, which is based on an assumption of homogeneity (upper panel of figure).  This rarely applies in practice but there are more sophisticated models which allow for different kinds of clustering or heterogeneity (lower panel of figure, generated in R using the ‘spatstat’ package).

Analysis of count data at LSHTM

Please contact Neal Alexander while we put together a more comprehensive list.  His work on count data includes quality control of malaria blood slides (→ Measurement error and misclassification) and spatial analysis of parasite counts (→ Analysis of clinical trials including prognostic models and Design and analysis for dependent data).

Comments are closed.