Ruth Tiffany Naylor, I'm a bit worried about there being nine factors in the original data (if I understand you correctly). To me, having so many factors suggests that the factor analysis back at the start was not conducted properly. For example, if the Kaiser criterion was used (i.e., accepting that the number of eigenvalues > 1 is an indicator of the number of factors in the data), it is likely that too many factors had been extracted.

However, I'm not sure what the original setup was - i.e., how nine factors were believed to exist - so my concern might be a bit off beam.

Is there any way you could perform factor analysis on (your) fresh data? In my experience, different samples can easily have different factor structures in their data, and it could be misleading to conduct analyses that assume the existence of factors that don't, in fact, exist.

Ruth Tiffany Naylor If you can find a gap in extant literature.

Good luck.

It would probably help to conduct tests of longitudinal measurement equivalence (using longitudinal CFA) to check whether the rewording had any effects on the parameters in the measurement model (loadings, intercepts, measurement error variances). If these parameters changed substantially across time (meaning measurement equivalence does not hold), then the latent factor means and/or variances may not be comparable across time. In other words: You may be measuring different things across time.