[R-meta] Multiple time point meta-analysis in metafor
Daniel Quintana
daniel.quintana at medisin.uio.no
Thu Sep 14 12:50:12 CEST 2017
Hi all,
I have a question regarding a multiple time point meta-analysis that I would like to conduct using metafor.
For this meta-analysis of a treatment for psychiatric illness, I’m interested in two different treatment outcomes (Outcome A and Outcome B) at two time points (time 1 and time 2). Most studies in this area compare an active treatment with a placebo using a between-participants design, and then assess both Outcome A and Outcome B over two time points.
I have estimated the standardised mean difference (SMD) and variance for both outcomes and both time points for each study, which can be found in the following data frame:
> study <- c("study 1", "study 2", "study 3", "study 4", "study 5", "study 6", "study 7", "study 8", "study 9", "study 10")
> y1i.w <- c (0.332, -0.421, 0.129, 0.000, 2.836, 0.267, 0.114, 0.051, 0.207, 0.132)
> v1i.w <- c (0.043, 0.114, 0.034, 0.052, 0.078, 0.059, 0.027, 0.108, 0.030, 0.200)
> y2i.w <- c (0.231, NA, -0.199, 0.000, 0.933, 0.005, 0.255, NA, 0.136, NA)
> v2i.w <- c (0.065, NA, 0.034, 0.058, 0.048, 0.049, 0.033, NA, 0.030, NA)
> y1i.p <- c (-0.226, 0.386, -0.179, -0.170, -3.137, 0.069, 0.149, 0.120, NA, 0.663)
> v1i.p <- c (0.043, 0.114, 0.034, 0.053, 0.087, 0.058, 0.027, 0.108, NA, 0.211)
> y2i.p <- c (-0.191, NA, -0.199, -0.218, -1.550, -0.272, 0.130, NA, -0.106, NA)
> v2i.p <- c (0.065, NA, 0.034, 0.059, 0.056, 0.049, 0.033, NA, 0.030, NA)
> dat <- data.frame(study, y1i.w, v1i.w, y2i.w, v2i.w, y1i.p, v1i.p, y2i.p, v2i.p)
Here, y1i.w = SMD at time 1 for Outcome A, v1i.w = variance at time 1 for Outcome A, y2i.w = SMD at time 2 for Outcome A, v2i.w = variance at time 2 for Outcome A, y1i.p = SMD at time 1 for Outcome B, v1i.p = variance at time 1 for Outcome B, y2i.p = SMD at time 2 for Outcome B, v2i.p = variance at time 2 for Outcome B.
Ideally, I would like to investigate effects at each time point (Outcome A and Outcome B separately), compare effects between timepoints (Outcome A and Outcome B separately), and compare effects at each timepoint (Outcome A vs. Outcome B).
There’s no data available for within-study correlations, so I’m assuming this is 0.97. I’ve tried to follow the “dat.ishak2007" example in the metafor package documentation to create the V matrix, but I can’t figure out how to create this matrix for the dataset I have.
I would appreciate any assistance!
Kind regards,
Daniel Quintana
[[alternative HTML version deleted]]
More information about the R-sig-meta-analysis
mailing list