I have used type = complex to run a nested (restricted) model and a comparison model. The comparison model is exactly the same as the nested model except that it tests 4 additional parameters (a new IV predicting 4 existing DVs).
Despite the fact that my nested model estimates fewer parameters in total (48) than the comparison model (52), the nested model has fewer degrees of freedom (14) than the comparison model (17). Is this wrong? If so, why might this be the case - is there likely to be something wrong with my data or model?
Consequently (I assume), the Satorra-Bentler scaled chi-square difference test used to compare the fit of these models produces a negative chi-square difference value. Is this uninterpretable? If so, is there another way to compare model fit?