Hello,

I am in the process of conducting a latent transition analysis examining the stability and change of mental health profiles across three time points. I have already conducted separate LPAs for each measurement point, and multiple fit indices support a four-class solution for each time point. Examination of the means of the different scales supports four conceptually similar groups at each point.

While the groups are conceptually similar across time, the means are not necessarily the same across times, which is expected given that internalizing and externalizing symptoms tend to increase from childhood to early adolescence.

Is it necessary or possible to test for configural invarance statistically? Or would this not make sense, given expected developmental differences?

I am also having an issue with Mplus when examining the results of the LTA. While the same groups are identified at each point, the class number assigned to each group is different, which is producing strange results. For example, regressions predicing C2 membership from C1 membership are not significant, becuase the classes are in a different order. Is it possible to fix the order of classes when running an LTA in Mplus?

Any suggestions would be greatly appreciated.