Mediation analysis using multi level mixed effects models for longitudinal cluster randomized trial

[Andrea Skye]

Hello!

I would be very grateful for some advice please.. I created a crude model for my analysis analysing an intervention's effects on haemoglobin levels in children taking into account inter cluster variance at village and household level:

meqrlogit hb_c i.month##intv strat if month==1 | month==6 & sub_6==0|| vlligeid : || hhid:, or

where hb_c is the outcome variable for haemoglobin levels in children.

I am trying to analyse mediating variables and mediating effects by fitting this crude model into a sem command and I keep getting errors. Is there any other (less computationally intensive) method of analysing mediators in this type of study? I have already tried the medeff and gformula commands and they don't seem to be working..

Thank you!

A

[Roman Mostazir]

see the help file of Generalised Structural Equation Model help gsem for multi-level structural equation models. The non-linear paths for mediation effect can be tested in the usual way using -nlcom- command as it is in -sem- .