# How to perform statistical power analysis for generalized linear models on IBM's SPSS?

I ran a generalized linear model (Poisson distribution) on some data of mine. This is my simplified syntax:

{(Poisson) GLMM embryo count BY Lineage(Ploidy), Ploidy, P treatment,
Ploidy*P treatment, Length}


My dependent variable is embryo count (discrete data). My independent variables are qualitive excluding length (which is continuous data). Lineage is can be thought of as a family or tribe and ploidy is the number of sets of chromosomes my organism has.

When I perform a general linear model on SPSS it is relatively easy to do a power analysis since there is an option to check on a power analysis button and the user receives "Observed Power" values. I have no idea how to do a power analysis for a generalized linear model or if there is a more appropriate way to perform a power analysis on my general linear model. Via syntax or drop-down menu are both great options for me!

I just looked at my results and it seems like there isn't any type of pattern with the observed power.

• 'observed power" is not a power analysis. Well, it is but it's post hoc and doesn't tell you anything you didn't know (if your result is statstically significant, your power is high, your result is no significant, your power is low). Mar 1, 2022 at 17:57
• @JeremyMiles Thank you for that information. Do you have any information (links/pdfs) that I can read that provide information on how to do a power analysis? I have found this link, but other information would be helpful. Mar 9, 2022 at 20:31
• In SPSS specifically? Sorry, I don't know. Mar 10, 2022 at 14:17
• See this page for some details on how this is done. As @JeremyMiles says, this type of "observed power" isn't useful except perhaps as a guide to design of subsequent experiments. This page suggests that you need extra modules to do power analysis with SPSS, but that thread is now 10 years old. Russ Lenth's tool linked from there, many tools in R, and simulation are ways to do power analysis for prospective study design.
– EdM
Apr 23, 2022 at 14:31