Mutilevel model with time and cluster effects and mediation

[plf515]

I have a data set that is a bit complex.

People were part of various organizations (there were 80 organizations).  Organizations were randomly assigned to one of two treatments (so there is a cluster effect) and measured at three time points (so there is a time effect).  The main interest is in mediation and, because the data are quite non-normal, I need to bootstrap the effect sizes.

Any hints on how to approach this would be appreciated.

And, if you're going to SESUG 2016 - maybe we can discuss it there.

I am using SAS 9.4

[Ksharp]

It is longitudinal data (repeated measure). Check PROC GEE/MIXED/GLMMXI ......

@Steve @lvm can give you good advice .

[plf515]

Thanks K. Yes, PROC MIXED is probably the spot, but I am not sure if it can do this and if so, how it can.

[SteveDenham · Accepted solution]

Hi Pete @plf515r,

Someplace in the SAS-L archives there is a piece of code that David Cassell wrote that can be used as a wrapper for almost any kind of randomization test, which ought to include bootstrapping.  I think I might have something remaining of it  It's from SUGI 27.  Hope this helps.

Steve Denham

[plf515]

Thanks Steve. I will look for it



Peter

##- Please type your reply above this line. Simple formatting, no
attachments. -##