tidy_multgee Incorrectly Defines Values in y.levels for Nominal Logistic Regression

[jackmwolf]

tidy_multgee defines the variable y.levels, which, I believe, are intended to correspond to the levels of the response associated with each covariate. However, these levels are currently defined by taking factor levels (2, 3, ..., J). However, multgee::nomLORgee is parameterized such that factor level J is the reference group and covariates are given for the odds of levels (1, 2, ... J-1) versus the reference group (see the fitted marginal baseline category logit model in ?nomLORgee). Therefore, y.levels should correspond to the first J-1 levels of the factor, and not the last J-1 levels.

[larmarange]

Thank you for having spotting it and thanks for the PR. I'm currently traveling but will look at it in the following weeks.