I am handling an analysis where I have count data for a set of samples before and after a treatment. I wanted to fit a Poisson regression using the Fit model Platform, but I am not sure if and how I can include information for the sample pairing.
Also, my data have around 20% of 0 counts. Is it worth contemplating Zero-inflated options?
Some more information might help, but a generally preferred method of analyzing pretest-posttest scores or counts is an Analysis of Covariance, or ANCOVA.
Are your pretest counts comparable? Are they correlated with the posttest counts? Does Regression To The Mean influence posttest counts?
Without more information, this is kind of a shot in the dark, but try setting up your Fit Model with the posttest count as the Response and the pretest count as a covariate. Contemplating a ZIP model might also be helpful. JMP documentation has some great examples of ANCOVAs.
I am not sure I can use ANCOVA since I do not have a two (or more) level factor. I have an experiment where I measure number of particles on different positions of a surface before and after cleaning the surface. My goal is to see if the cleaning method has reduced the number of particles on the surface. There is a trend indicating that after cleaning, each position shows a lower count of particles