I am running a binomial logistic regression in R. My colleague is running the same regression in SPSS. Our DV is dichotomous and data is non-normally distributed. The IVs are a mix of categorical and continuous.
When I compare our outputs, I notice that our estimate coefficients, standard error, and p values are the same. However, I get a z value in R, where she gets a Wald value. I have been reading and think that these are similar/the same with the exception that the Wald does not assume normality where the z does.
First question: Do I understand this relationship correctly?
Second question depending on answer to first question: If so, is there a way to compute the Wald statistic in R to assume non-normality when running the model?
If not, why would our outputs differ in this manner?
Example of what differs in our outputs:
Coefficients: Estimate Std. Error z value Pr(>|z|) (Intercept) 2.48462 1.17860 2.108 0.0350 * age -0.02332 0.01087 -2.145 0.0319 * edu1 0.76923 1.16412 0.661 0.5087 edu2 0.60235 1.12535 0.535 0.5925 edu3 1.23647 1.13438 1.090 0.2757
B S.E. Wald df Sig. Exp(B) age -.022 .011 4.270 1 .039 .978 edu .296 .201 2.173 1 .140 1.344 Constant2.661 .513 26.905 1 .000 14.314
Note: I understand the "scripting" nature of this question (R vs SPSS). However, I feel that the question more concerns the analytic output and the Wald/z statistic rather than the specifics of the programs.