Treatment-covariate interaction in ipdmetan!
Hi everyone,
I'm running a two-stage IPD meta-analysis by using ipdmetan to assess treatment-covariate interaction (e.g., group [LBP vs control] x BMI covariate) on postural control measures.
I would like to ask whether is possible to NOT include a full factorial interaction, for example:
(1) ipdmetan, study(StudyID) interaction random(reml) effect() title(RMS-EC-AP - Group × Age) texts(110) forest(favours(Control # LBP)) xlab(-6 0 6): xtmixed RMS i0.GroupTwo#c.BMI i0.GroupTwo i.Sex c.Age if Eyes==1 & Direction==1
Here in this model, I only included i0.GroupTwo#c.BMI i0.GroupTwo, but I did not include c.BMI
Or should I use a full factorial interaction:
(2) ipdmetan, study(StudyID) interaction random(reml) effect() title(RMS-EC-AP - Group × Age) texts(110) forest(favours(Control # LBP)) xlab(-6 0 6): xtmixed RMS i0.GroupTwo##c.BMI i.Sex c.Age if Eyes==1 & Direction==1
Here in this model, I included i0.GroupTwo#c.BMI i0.GroupTwo c.BMI
This because the results in some measures were significantly different between both models. The first model indicated a significant difference with higher effect size while the second one showed no significant difference.
Thanks
0 Response to Treatment-covariate interaction in ipdmetan!
Post a Comment