I am running a conditional 'linear' latent growth curve model. Does it matter how the categorical predictor variables are coded (i.e., using 0/1 or 1/2 code)? I ask because I seem to get vastly different results when I code them differently?
Secondly, my unconditional model showed a significant negative slope - outcome decreased over time. When I add predictors to the model, the mean of the slope is no longer significant (although there is still significant variability around the slope mean). Why would this be?