0
$\begingroup$

This is my test design which is simply a grouping variable:

enter image description here

When I choose to regress a gender variable (w/m) onto this grouping variable with a multinomial logisitic regression model I'll observe

enter image description here

Why does SPSS list both categories of gender with one being redundant rather then choosing one by default (the last category) as the reference?

If there's no reference for the gender variable the interpretation of the intercept gets rather confusing. Is there simply a mistake involved or do I miss something?

$\endgroup$
2
$\begingroup$

Gender = w is the reference category. It has B = 0 in all cases.

$\endgroup$
2
  • 2
    $\begingroup$ +1. In other words, SPSS is actually behaving as per @Druss2k's initial expectations -- and the way this information is conveyed in the parameter estimates is with a zero estimate for beta, with no associated standard error. Other software sometimes does similar things in this respect (e.g. Proc Genmod in SAS, from memory) $\endgroup$ Feb 2 '13 at 22:58
  • $\begingroup$ Ah ok, I'm new to SPSS and in R the reference category is not displayed hence my confusion. But I find this kind of display rather inconsistent with other output SPSS provides since the reference category in a logistic regression model is not displayed in the same way. In this case the reference category is contained in the intercept estimate. $\endgroup$
    – Druss2k
    Feb 2 '13 at 23:32

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.