I am running a multinomial logistic regression equation (DV has 3 categories). One of my independent variables has 3 categories. The overall test statistic is not significant (p=.1223). However, when I look at the analysis of maximum likelihood estimates, one of the comparisions is significant (p=.0357).

My question is, do you have to have an overall significant test statistic in order to interpret the estimates? And, why is it that the overall test isn't significant but significant differences between categories emerge? Thanks in advance to anyone who can help.

Meghan