I am running a growth model looking at trajectories of conflict over time. I have a pair of dichotomous time invariant predictors that I want to use in my model.
When I inclue the Var1*time and Var2*time interactions in the model I can easily graph the growth trajectories by reverse coding the variables and recentering the time variable (3 time points: 0, 1, 2; -1, 0, 1; -2, -1, 0). When I re-run the analysis and different centerings of time, the coeffiecient for time stays the same (save for when I reverse code the variabes).
Ok, so far so good....
The problem is that when I specify a three way interaction (Var1*Var2*time) and I re-run the analysis with recentered time, the coefficient for time changes. This should not happen and only happens when I have the 3-way interaction.
Why is the coeffiecint changing? It seems like it is supposed to change but I cannot for the life of me determine why. Any input would be much appreciated!!!