Longitudinal analysis of drug use data

[omnibus]

Hi all,

I'm looking to conduct a longitudinal analysis using data from 5 waves of an observational research study. What I'm really interested in doing is seeing how changes in parental monitoring and some other family factors (i.e., cohesion, etc.) impact drug use and negative consequences of drug use.

I had originally thought about running a mixed effects model/growth curve analysis. The problem I'm having is that while I have data on the family variables and drug use for all 5 time points, I only have data on the my second outcome variable - negative consequences (a summed score of the AUDIT measure) at the last 3 time points.

Does anyone have suggestions on the appropriate longitudinal technique that could be used in this scenario?

 

[Disvengeance]

Are your outcomes continuous or categorical?

I believe that missing the outcome variable for the first two time points would affect the interpretation of your results rather than the method of analysis per se, but I could be wrong.

 

[omnibus]

Yes, I'm just wondering if there is a more appropriate technique I'm not aware of that would be worth looking into.

 

[Lazar]

I wonder if you could adjust the coding of time here i.e., for the variables you have all 5 waves for code as -2,-1,0,1,2 and for those with only three code as 0,1,2. That way the intercept would be common for both variables. Depending on the specifics of you question, however a autoregressive crosslag model might be easier to deal with.

EDIT: Obviously this coding scheme would only work in a structural equation modelling approach to growth curves.

 

[omnibus]

That's an interesting idea. I'll give it a shot. Thank you for the suggestion!