0
$\begingroup$

I am conducting a meta-analysis of studies that didn't use a control comparison group. I use metafor on R. I do have the R code for studies with a control but not without a control. My outcome is continuous (0-10 scale). Any help would be much appreciated it. Thank you so much, Mina

$\endgroup$
  • $\begingroup$ Can you say more about what your outcome represents? If it's a mean score across participants in a given study, do you have standard deviations and sample sizes corresponding to each mean? $\endgroup$ – Emily May 30 at 14:54
  • $\begingroup$ Hello, thanks for your comment. yes the following are the measures I have:m_pre; m_post; sd_pre; sd_post; ri (calculated) $\endgroup$ – Mina Lazaridou May 30 at 18:37
1
$\begingroup$

You will want to use a mean difference standardized by either the raw score (measure = "SMCR") or the change score (measure = "SMCC"). Raw score standardization involves dividing the difference between the time 1 and time 2 means by the standard deviation of the time 1 scores, while change score standardization involves dividing by the standard deviation of the change scores.

Which one you want to use depends on a couple things.

Is your meta-analysis testing differences across alternate treatments or experimental conditions? If so, the raw score standardization is preferable. If your meta-analysis is primarily concerned with changes in individuals, the change score standardization is better.

Does the test-retest correlation differ substantially across studies? If so, the raw score standardization is preferable, because it doesn't depend on that correlation. In contrast, the effect size based on change score standardization will be increased when the test-retest correlation is high and decreased when the test-retest correlation is low.

See more on this page about how to write the R code to calculate the effect sizes, under Outcome Measures for Change or Matched Pairs -> Measures for Quantitative Variables.

I also highly recommend reading the following paper on this topic: Morris & Deshon (2000).

$\endgroup$
  • $\begingroup$ Thank you so much Emily. This is the syntax I used before data <- escalc(measure="SMCR", m1i=m_post, m2i=m_pre, sd1i=sd_pre, ni=N, ri=ri, data=data) I wasn't sure if this is correct. From what I read I had to also run the following for the actual meta-analysis but this gives me the odds ratio which is not correct for pre-post differences without a control group: ma_model<-rma(yi, vi, data=data, method="FE", digits=2) $\endgroup$ – Mina Lazaridou May 31 at 22:44
  • $\begingroup$ The syntax for calculating the effect size looks right to me. I would recommend posting a separate question where you ask about why it's giving you an odds ratio, including the code you used, the output you got (indicating where you're seeing odds ratios), and some sample data. $\endgroup$ – Emily Jun 1 at 3:39
  • $\begingroup$ Thank you so much. I thought the output is odds ratio because the total effect size is -2.25 (CI: -2.39, -2.10). I am used to cohen effect size and this is my first time using metafor that's why is confusing to me.thanks again for your input! $\endgroup$ – Mina Lazaridou Jun 1 at 15:50

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.