I am running a parallel process growth model with a time-invariant moderator, but have difficulties on the interpretation. My DV is well-being, my IV is friendship quality (both measured at 3 time points) and have an observed moderator (ethnic identity) measured at T1.
I am intersted to see whether the initial levels of friendship quality and ethnic identity and their interaction (defined by the XWITH command) predict the slope of well-being.
Results showed a negative effect of quality intercept, a negative effect of identity and a positive interaction on well-being slope.
1) I guess the negative main effects on well-being slope do not show that initial levels of quality and identity lead to decreases in well-being, but rather means that higher initial levels of quality and identity lead to lower levels of change in the slope of well-being? But how do I understand the direction of the change?
2) How can I interpret the positive effect of interaction?
3) For plotting the simple slopes between a latent growth factor and an observed variable, how can I obtain the variances of the IV, moderator and the interaction?
Thanks Linda, also just a basic question to confirm:
My univariate models showed that friendship quality increases over time and well-being is stable over time.
If I find a significant and positive covariance between these two slopes, does this mean that increases in quality is associated with the increases in well-being, or increases in quality is associated with more stable well-being?