0

I wish to calculate my effect size based on the response ratio but my studies only have means of the treatment(no-tillage) and control(conventional tillage). However, my studies didn't provide variance or standard deviations. How do I go about this especially with metafor. I get this error below

dat <- escalc(measure="RR", m1i=NT,m2i=CT, data=dt) Error in escalc(measure = "RR", m1i = NT, m2i = CT, data = dt) : Cannot compute outcomes. Check that all of the required information is specified via the appropriate arguments.

Acha
  • 1
  • 1
  • Welcome to stackoverflow, please provide sample code along with data what you have tried. – Arun kumar mahesh Jul 10 '20 at 05:09
  • Hi @Acha, please have a look at the following link which provides the detail about how to produce reproducible example. https://stackoverflow.com/questions/5963269/how-to-make-a-great-r-reproducible-example – monte Jul 10 '20 at 05:53

1 Answers1

0

measure="RR" is for (log) risk ratios. If you want response ratios (i.e., log transformed ratios of means), you need measure="ROM". But you also need the sample sizes and standard deviations of the two groups. Otherwise, the sampling variances cannot be computed. See help(escalc) and see help(dat.curtis1998) for an example.

Wolfgang
  • 2,810
  • 2
  • 15
  • 29
  • Thank you Prof Wolfgang. Though the studies selected didn't report sample sizes and standard deviations(SDs), I can compute my sample sizes from the study using experimental plots and replicates however for SDs, is it logical to calculate the SD for each group based on the extracted means from the studies? or must the SD be based on those extracted from each particular study? – Acha Jul 10 '20 at 20:24
  • You really need the SD of the measurements within studies. The SD of the means is something completely different. – Wolfgang Jul 10 '20 at 22:46