0

I am having a mighty difficult time interpreting the following results given to me based on a generalized linear model with a Poisson distribution: enter image description here

Data: data represents the number of visitors to a store and whether their visits were necessary (they really needed something) while they were store members or not. Comparing stores 1 and 2.

Question 1: I am trying to interpret the meaning IRR here. What does IRR of 0.77 mean for Unnecessary category? Also, what does IRR of 1.12 mean (since it's above 1).

Question 2: so given all z-values are below 2, does it mean none of the results are statistically significant?

kjetil b halvorsen
  • 63,378
  • 26
  • 142
  • 467
BitwiseStarter
  • 25
  • 5
  • Have you tried searching the site for [interpreting poisson](https://stats.stackexchange.com/search?q=interpreting+poisson)? – Alexis Feb 18 '18 at 17:49
  • @Alexis I've read not only every single answer on this site but also many, many other sources: https://onlinecourses.science.psu.edu/stat504/book/export/html/165 – BitwiseStarter Feb 18 '18 at 19:08
  • https://www.statsdirect.com/help/regression_and_correlation/poisson_regression.htm – BitwiseStarter Feb 18 '18 at 19:08
  • and this one: http://pdf.th7.cn/down/files/1508/Regression%20Analysis%20by%20Example,%205th%20Edition.pdf these are just tip of the iceberg. – BitwiseStarter Feb 18 '18 at 19:09
  • closest one you get on this site is this: https://stats.stackexchange.com/questions/314068/interpreting-incidence-rate-ratios-in-poisson-without-exposure-variable but if you look at it, nobody has provided an answer. So if you know it, post an answer, it'll benefit me and everyone else, including about 20 of these questions posted without answers. Doubting people's due diligence on posting questions won't benefit anyone. – BitwiseStarter Feb 18 '18 at 19:11
  • Your question would be improved by editing to add what you do and do not understand about answer to other questions about interpreting Poisson regression. For example, are you asking a question about your model specifically, or about the meaning of IRR? Your question about statistical significance is currently meaningless since (a) you have not defined your preferred type I rejection rate ($\alpha$), and (b) you have not defined whether that preferred rejection rate is with respect to a single test, a familywise error rate, or a false discovery rate (i.e. w/r/t multiple comparisons). – Alexis Feb 18 '18 at 19:26

0 Answers0