I need to carry out a Hausman-McFadden test before I run a multinomial logistic regression on my data. While I've successfully run an Hausman-McFadden test on example data and probably could for my actual data, I'm still a little confused about what precisely I run the test on.

For clarity, my data currently stands as 9 variables total, with 8 continuous, dichotomous and categorical variables contributing to 1 categorical dependent, which is whether a levee on a river is passable by migrating fish (5 levels).

I don't understand what to run the test on out of my 9 variables. Do I do it for every single one? Do I assess every possible combination (because that will literally take days...?) or do I only need to compare one choice to the rest and thats it? ANY clarification or links to useful advice (please bear in mind that i've been learning R/stats for about 2 weeks and my knowledge is unfortunately very limited!) would be much appreciated, thanks!