This is what I'm trying to do in S-Plus, using the GUI interface.
Again, here's the SAS code:
Data meta;
Infile 'c:\meta.txt';
Input effect sd week studyID;
Invsd=1/(sd);
Run;
Proc mixed covtest data=meta;
Class studyID;
Weight invsd;
Model effect = week /s;
Random int /type=un sub=studyID;
Parms (0.1) (1)/hold=(2);
Run;
And here's what I'm doing in S-Plus GUI.
First, I load the spreadsheet into the GUI. The data is in columns, in the
following order:
studyid effect sd week invsd
I go up to "Statistics", then choose "Mixed effects", then "Linear."
Data set is "Raudenbush"
In "Fixed effects", I choose effect as the dependent. For "independent", I
choose week. This results in the formula:
effect~week
In "Random effects", I choose "studyid" as a group variable. This creates
the random formula ~1 | studyid
I then click "Options."
I choose "Within group variance" as "Fixed."
I choose "Invsd" as the covariate since this is the weight variable.
I choose "Within group correlation" as "General Symmetric"
Group variable as "studyid"
Not sure what to put for parameters...I put "1" because I want to hold the
lowest covariance structure as with the SAS code, but I don't know if I'm
doing this correctly.
When I run the stats, I get an error message that says:
"Problem in varFixed: argument form= not matched: varFixed(form = ~invsd, )
I can't seem to figure out how to implement the above SAS code into
equivalent S-Plus code.
James
----- Original Message -----
From: "jagadish rangrej" <jrangrej@yahoo.com>
To: "James Krieger" <jkrieger2@cox.net>
Sent: Saturday, February 28, 2004 2:24 PM
Subject: Re: [S] How to set up 2-level hierarchical meta-analysis using
GUIS-Plus for Windows?
> Hello James,
>
> I had been using BUGS for fitting such hierarchical
> model but perhaps I use to fit Bayesian Hierarchical
> models.
>
> I can see that you shown "proc mixed" from SAS, then
> definitely you can go and use lme from (nlme package)
> of R/Splus, I'd suggest you to go with lme package in
> R.
> That is quite extensive package which can be used for
> building such hierarchical models. But note that these
> under the assumption that error structure is normal in
> both the cases.
> hope that helps,
> Jagadish.
> --- James Krieger <jkrieger2@cox.net> wrote:
> > I am currently doing a meta-analysis of studies that
> > look at the effects
> > of protein intake on body composition. I want to
> > set up a 2-level model
> > that models the within and between study variation,
> > as outlined in Joop
> > Hox's chapters on meta-analysis in his books. I've
> > seen examples for HLM and SAS on the
> > web, but these are not the pieces of software that I
> > use. In an attempt
> > to learn how to do a meta-analysis with the GUI
> > interface of S-plus (I'm also trying it with
> > GenStat..I've cross-posted this problem to the
> > GENSTAT list so I hope that doesn't bother anyone),
> > I've
> > tried to adapt the following SAS code (for a
> > meta-analysis of teacher
> > expectancy effects by Raudenbush):
> >
> > Data meta;
> > Infile 'c:\meta.txt';
> > Input effect sd week studyID;
> > Invsd=1/(sd);
> > Run;
> >
> > Proc mixed covtest data=meta;
> > Class studyID;
> > Weight invsd;
> > Model effect = week /s;
> > Random int /type=un sub=studyID;
> > Parms (0.1) (1)/hold=(2);
> > Run;
> >
> > I've been trying to do something similar in both
> > S-Plus 6.2 and GENSTAT 7.0 without
> > success.
> >
> > Anyone who might be able to help me with how to set
> > up a multilevel meta-
> > analysis with S-Plus 6.2, it would be tremendously
> > appreciated.
> >
> > Thank you for your time,
> >
> > James Krieger, M.S.
> > Graduate Assistant, Nutrition
> > University of Florida
> > Webmaster, WSU Strength and Conditioning
> > http://www.wsu.edu/~strength
> > Science Editor, Pure Power Magazine
> > http://www.purepowermag.com
> >
> >
>
>
> __________________________________
> Do you Yahoo!?
> Get better spam protection with Yahoo! Mail.
> http://antispam.yahoo.com/tools
>
|