Longitudinal Analyses with questionnaire with varying items per wave

#1
Hi all,

For my thesis I am conducting a longitudinal analyses (5 waves) on antisocial behavior measured with the ASBQ. Now I have a different amount of items per wave, because they included items that were more age appropriate per wave. This results in: T1 (31 items), T2 (26 items), T3 (28 items), T4 (29 items), T5 (29 items). In total 18 items are measured across all 5 waves.

I planned to do factor analyses and examine measurement invariance across time, and thus constrain the factor loadings etc to see whether my latent construct would mean the same over time. But now I see items vary across waves, I am unsure wheter I can still keep the items in the analyses that are unique to one wave. Should I only keep the 18 items that are present in all the waves?

Thank you!
 
#2
my suggestions-
#1 let theory guide your choice on whether to include only the 18 items present in all instruments or the expanded instruments

#2 here is how to provide statistical evidence that inclusion of more items does not threaten statistical inference:
It is similar to testing for model invariance. In this case, you want to extract a stability coefficient using an auto regressive model coupled with testing for invariance. If the stability coefficients are statistically the same (i.e the both models are stable over time in respect to each other) and they are invariant, then they are statistically the same.