[R-meta] Multiple use of the same control

[Viechtbauer, Wolfgang (SP)]

No, this accounts for auto-correlation in the underlying true effects, not the sampling errors. For the latter, you have to construct an appropriate V matrix.

This is why I am asking. If you are not dealing with the autocorrelation in the sampling errors due to repeated measurements in the same group over time, then trying to account for the dependency in the sampling errors due to reuse of control group data is only tackling half of the problem.

I can already tell you though that the construction of a V matrix that accounts for both of these sources of dependency is likely to be very difficult. A recourse would be to ignore the dependency in the sampling errors altogether and then use cluster-robust inference methods.

Best,
Wolfgang

-----Original Message-----
From: Arne Janssen [mailto:arne.janssen using uva.nl] 
Sent: Thursday, 08 November, 2018 17:35
To: Viechtbauer, Wolfgang (SP)
Cc: 'r-sig-meta-analysis using r-project.org'
Subject: Re: [R-meta] Multiple use of the same control

Dear Wolfgang,

Thanks for your reply. As far as I understood your documentation at 
https://wviechtb.github.io/metafor/reference/rma.mv.html, I can do this 
by using struct="CAR", is that correct?

Best wishes,
Arne

On 08-Nov-18 16:42, Viechtbauer, Wolfgang (SP) wrote:
> Dear Arne,
>
> Are you allowing for correlation between the sampling errors over time?
>
> See, for example:
>
> Musekiwa, A., Manda, S. O., Mwambi, H. G.,&  Chen, D. G. (2016). Meta-analysis of effect sizes reported at multiple time points using general linear mixed model. PLOS ONE, 11(10), e0164898. doi:10.1371/journal.pone.0164898
>
> See also: help(dat.ishak2007) and help(dat.fine1993).
>
> Best,
> Wolfgang
>
> -----Original Message-----
> From: R-sig-meta-analysis [mailto:r-sig-meta-analysis-bounces using r-project.org] On Behalf Of Arne Janssen
> Sent: Saturday, 03 November, 2018 16:32
> To: 'r-sig-meta-analysis using r-project.org'
> Subject: [R-meta] Multiple use of the same control
>
> Dear list members,
>
> I have a similar problem as Lasse Bang ( Follow-up - Dependencies in
> data: multiple studies with overlapping sample sizes).
>
> I want to do a meta-analysis of studies with  time series, often with
> time series of several treatments and one control time series. To
> include all the treatments in the analysis, I therefore need to use the
> same control time series several times. How can I correct for that?
>
> I am using  rma.mv with time as one of the modulators and StudyID (the
> publication ID) and SampleID (time series within the same study) as
> random effects, and a continuous-time autoregressive structure.
>
> Thanks in advance for any advice.
>
> Best wishes,
> Arne Janssen