So I've been running a few ordered logit models in R using the polr() function and I've run into an error that I'm not familiar with.
My dependent variable is factor variable with three levels. All of my independent variables are numeric and three of them are binary variables. It's only one independent variable that's giving me trouble. I have a variable for tuition that I rescaled because I had been running into errors with other variables that had a really large max. The new rescaled tuition variable has a range of 5.39-42.56. When I run the polr() function with everything else it runs fine with no issues. Then when I add tuition to the model I get the following error message:
Error in optim(s0, fmin, gmin, method = "BFGS", ...) : initial value in 'vmmin' is not finite In addition: Warning message: glm.fit: fitted probabilities numerically 0 or 1 occurred
When I run the model with just the tuition variable and my dependent variable, it runs fine. I've checked the correlations for all my variables and none of them are close to 1.