I am attempting to run mixed effects models (using the REML method) in the fit model platform in JMP 8.0.1. I have two random factors and two fixed factors. I have run the same model for two different response variables (so the structure, number of reps etc is identical) and am puzzled by the fact that the reported denominator DF in my fixed factor effect test report is not the same for both models, and the numbers are extremely strange for one model where I am getting DFDen for my treatment of interest as 4.109 in one model and 1691.0 in the other (my total number of observations is 1757). Can anybody explain to me why this would be the case, or how if I might be running the model incorrectly?

Thanks.