Elisa posted on Sunday, December 18, 2011 - 7:16 am
I have a longitudinal factor model with 9 observed variables measured at 3 time points (3 per time point). I checked for measurement invariance and found out that some intercepts vary over time. Now I'm really struggling to interpret them! For example:
Intercept 1 is 28 at Time 1 and 29 at Time 2: Does this mean that participants find it easier or more difficult to agree with this measure? (a higher score for this item indicates more problems).
The behavior of the intercept depends on how you treat the factor means. The mean of an item is a function of the intercept, the loading, and the factor mean, so if the mean increases over time it doesn't imply that that the intercept needs to increase.
Your question is best answered if you have intercept and loading invariance over time for 2 of your 3 items, but not intercept invariance for the third item (let's assume loading invariance for that item for simplicity), and you fix the factor mean to zero at the first time point and free it at the other time points. In that scenario a higher intercept at time 2 implies that a subject at a certain factor value is likely to report more problems at time 2 than at time 1.
Elisa posted on Sunday, December 18, 2011 - 12:36 pm
Thank you very much for the prompt response! Very helpful!