[R-meta] How to conduct a meta-analysis on multiple-treatment studies with a repeated measure designs?

[Michael Dewey]

Dear Koenraad

Because you posted in HTML your data got completely mangled, at least 
when it arrived here. Perhaps you could re-post using plain text? Using 
dput() may help to ensure that readers get the same data-set as you sent.

Michael

On 14/05/2018 16:15, Koenraad van Meerbeek wrote:
> Dear all,
> 
> We want to do a meta-analysis with the metafor package in R to study the effect of multiple experimental treatments on species diversity over time. First of all, we use data from multiple-treatment studies, in which the effect of different treatments are compared against a single control group. So far, this is the same as the Gleser & Olkin (2009) example on the metafor website. However, some of the studies also measured the effects of the treatments over time (repeated measures design).
> 
> This is an example of how our data looks like (simplified). We also want to include magnitude of the treatment and duration of the study as moderator variables.
> 
> Study
> 
> Treatment
> 
> Year
> 
> Species diversity
> 
> Magnitude
> 
> Duration
> 
> 1
> 
> Control
> 
> 1
> 
> 1.35
> 
> 0
> 
> 1
> 
> 1
> 
> TR1
> 
> 1
> 
> 0.78
> 
> 0.75
> 
> 1
> 
> 1
> 
> TR2
> 
> 1
> 
> 0.23
> 
> 1.50
> 
> 1
> 
> 1
> 
> Control
> 
> 2
> 
> ...
> 
> ...
> 
> 2
> 
> 1
> 
> TR1
> 
> 2
> 
> 
> 
> 
> 
> 2
> 
> 1
> 
> TR2
> 
> 2
> 
> 
> 
> 
> 
> 2
> 
> 2
> 
> Control
> 
> 1
> 
> 
> 
> 
> 
> 1
> 
> 2
> 
> TRa
> 
> 1
> 
> 
> 
> 
> 
> 1
> 
> 2
> 
> TR2b
> 
> 1
> 
> 
> 
> 
> 
> 1
> 
> 2
> 
> Control
> 
> 2
> 
> 
> 
> 
> 
> 2
> 
> 2
> 
> TRa
> 
> 2
> 
> 
> 
> 
> 
> 2
> 
> 2
> 
> TR2b
> 
> 2
> 
> 
> 
> 
> 
> 2
> 
> 
> We started to calculate the log response ratio:
> dat <- escalc(measure = "ROM", n1i = dat$n1i, n2i = dat$n2i, m1i = dat$m1i,  m2 = dat$m2i, sd1i = dat$sd1i, sd2i = dat$sd2i)
> 
> And then fitted following mixed effects model:
> res.mv <- rma.mv(yi, vi, mods = ~ Magnitude + Duration, random = ~ Study| ID, data=dat)
> 
> We did not yet try to calculate a variance-covariance matrix as the Gleser & Olkin (2009) example, because we did not know how to take the repeated measures design into account.
> 
> How do you suggest to proceed? Expand the res.mv with a variance-covariance matrix? How would you do that? Or aggregate the data (across years) in some way and then follow the Gleser & Olkin (2009) example?
> 
> Best,
> 
> -----
> Koenraad Van Meerbeek
> Postdoctoral researcher
> Center for Biodiversity Dynamics in a changing world (BIOCHANGE)
> Section for Ecoinformatics & Biodiversity
> Department of Bioscience | Aarhus University
> Ny Munkegade 114, 8000 Aarhus C, Denmark
> E-mail: koenraad.vanmeerbeek[at]bios.au.dk
> Mobile: +32 479 206957
> 
> 
> 	[[alternative HTML version deleted]]
> 
> _______________________________________________
> R-sig-meta-analysis mailing list
> R-sig-meta-analysis at r-project.org
> https://stat.ethz.ch/mailman/listinfo/r-sig-meta-analysis
> 

-- 
Michael
http://www.dewey.myzen.co.uk/home.html