Negative Binomial vs Poisson GLM to Model Overdispersed Count Data
Summary
Here is a simulation to create synthetic overdispersed count data and to investigate how overdispersion effects the estimation of Poisson model standard errors. Additionally, example datasets are modeled with a negative binomial model (NB-2 variance parameterization) to investigate how the negative binomial model can more accurately model overdispersed count data.
Index
- overdisp_sim.R: contains helper functions used to create plot of coverage probability by overdispersion parameter.
- nbreg_examples.R: contains code for generating output for application datasets.
- CSV Files: datasets to be used in example applications.
- overdisp_analysis.html: the Rmarkdown file in HTML format.
Display the final writeup in a Rmarkdown file, using BitBucket HTML Preview, by clicking here.