This is my first time asking a question here, I have two sets of count data that I want to compare to one another. One set has an excess amount of zeros, while the other does not. My first idea was to use a Zero-Inflated Poisson regression for the first set of data with the excess amount of zeros within it, and then either a normal or quasi Poisson depending on the second set depending upon the existence of over-dispersion in the data. Is this a good approach or would it be better to use one analysis for both sets of data?
Thanks in advance.

  • $\begingroup$ What do you mean by "compare to one another"? Do you want to look at shape of distribution, measures of location/central tendency? Particular quantiles? Or what? $\endgroup$ – Peter Flom Mar 9 '16 at 20:06
  • $\begingroup$ I am sorry I want to determine if similar associations exist between the data. I am looking at two similar diseases and comparing it to the same environmental features. $\endgroup$ – B. Ashraf Mar 10 '16 at 5:35

So, from your comments, you have two diseases and you want to compare how they relate to a bunch of environmental variables.

If this is right, then you can simply use logistic regression on both data sets. There is no need for a count model because the dependent variable is not a count, it's a probability.

| cite | improve this answer | |
  • $\begingroup$ Your last statement is mysterious. The OP clearly indicates the dependent data are counts. In what sense, then, are you claiming these are "probabilities"? $\endgroup$ – whuber Mar 10 '16 at 13:55
  • $\begingroup$ Reading the OP's reply to my comment on the post, it's clear that the question was confused. OP does not want to compare the distributions but the regressions. $\endgroup$ – Peter Flom Mar 11 '16 at 2:55

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.