R help - May 2004 - Meta-Analysis using lme

Dear list-members,

I am trying to use R to conduct a meta-analysis, i.e. I'd like to use a
multi-level model to integrate the findings of a number of primary research
studies. 
 
I set up a simple two level-model (only summary statistics are provided by
each study) as follows:

sapp.lme <- lme(D ~ 1, data = sapp.frame, random = ~ 1 | STUDYNR, 
weights=varFixed(~-1+STDERR_D),na.action = na.exclude)

The intercept is random on both levels and the variable stderr_D (the
sampling variance) is supposed to be random only on level one. Besides, I
need to constrain the variance of stderr_d to equal 1. 

Could anybody help me correct the code I provided? Thanks in advance for
your time.

Cheers, 
Heike Heidemeier



--

On Tue, 11 May 2004, Heike Heidemeier wrote:
> Dear list-members,
>
> I am trying to use R to conduct a meta-analysis, i.e. I'd like to use a
> multi-level model to integrate the findings of a number of primary research
> studies.
>
> I set up a simple two level-model (only summary statistics are provided by
> each study) as follows:
>
> sapp.lme <- lme(D ~ 1, data = sapp.frame, random = ~ 1 | STUDYNR,
> weights=varFixed(~-1+STDERR_D),na.action = na.exclude)
>
> The intercept is random on both levels and the variable stderr_D (the
> sampling variance) is supposed to be random only on level one. Besides, I
> need to constrain the variance of stderr_d to equal 1.
>
I think someone posted a way of doing this some time ago, but in this case
I would recommend the rmeta package

sapp.meta <- meta.summaries(D,STDERR_D, method="random",
data=sapp.frame)

which will also do forest plots of the results.


	-thomas