I am testing a PPM. I first modeled separate LGMs for each construct and tested for gender differences in each LGM. Construct A showed a significant positive linear slope as well as a significant negative quadratic slope (I have 5 time points). Multi-group analysis of Construct A's LGM did not reveal any gender differences at this point. Similar results occurred for Construct B.
I then tested a general PPM of these constructs and found good model fit as well as significant connections between different growth parameters of the constructs. I then ran multi-group analysis to the PPM to investigate gender differences on the associations between Construct A and Construct B. Two group differences at the parameter mean level appeared: the linear and quadratic slopes for Construct A were significantly different for men and women - these parameters were significant only for men and not for women. Here is my problem: the direction of the slopes for Construct A for men in the PPM differ from the general LGM for Construct A - in the PPM, Construct A showed a significant NEGATIVE linear slope as well as a significant POSITIVE quadratic slope. The slopes for women, though non-significant in the PPM, follow the direction originally revealed in the LGM. I am not sure what this means - is my model unstable?
Thank you for any insight you may be able to provide me.