Count Regression with Sampled Count Variable
S. Jalil Kazemitabar · arXiv (Cornell University) · 2014
A common assumption in count models take the count variable as observable. I take it as produced via a sampling procedure with known sampling rate and reconsider the model. I assume multiple count models (describing the original count variable) and two different sampling procedure, random sampling with and without replacement. Under random sampling without replacement a Poisson count model of original count variable turns into another Poisson count model for the sampled count variable with the parameter multiplied by the sampling rate.