mediation analysis with longitudinal data - can I use change score as dependent variable?

Hey guys,

I'd like to run a mediation analysis in R with the following data:

  • "treatment" ("a","b","c") as predictor

  • "expectancy of relief" as mediator

  • "symptom intensity" as outcome variable

  • symptom intensity was measured at 3 times

  • t0: pre random allocation to the treatment groups as baseline

  • t1: midpoint

  • t2: postintervention

I expect treatment "a" to be superior to "b" and "c" in reducing symptom intensity. And I also expect that this effect is mediated by expectancy of relief.

Now, I have problems calculating the regressions for the mediation analysis:

For the predictor I have to code dummy variables I guess, right? But I'm not sure how to handle the outcome variable as I have longitudinal data. I thought of calculating a change score (t2-t0) and use it as the dependent variable. But I didn't find any papers really supporting this method even being bad as it leads to regression to the mean and stuff .. does anybody have a solution for me how I can handle the longitudinal outcome?

thank you very much in advance!!

This topic was automatically closed 21 days after the last reply. New replies are no longer allowed.

If you have a query related to it or one of the replies, start a new topic and refer back with a link.