Can I use the means difference (difference in mean between endpoint and baseline) and associated standard error to calculated a standard mean difference (SMD) and its 95%CI (I also have an exact p value and sample size)?
for context: it's for a meta analysis all of my other studies provided me with the means and SDs needed to get my SMD.
n=239, mean difference: −1.2 [SE 1.48]; p=0.4154
either formulas or online calculators or references are welcome :)
Thank you
I have not used it but here is an example of a calculator that has multiple functions. Please double-check if that works:
https://www.campbellcollaboration.org/escalc/html/EffectSizeCalculator-SMD1.php
Shin Murakami thank you, i used this link for my other calculations and it's mostly useful but they don't offer an option to use MD and SE.
OK, I assume you have baseline and endpoint measurements for each of the 239 subjects. I was lazy and used the Comprehensive Meta-Analysis (CMA) software to do the calculations for me. There are two possible ways you might calculate the standardized mean difference (SMD). The first one (see Test1.jpg below) is probably the best procedure. It makes use of your paired t-test p-value. You do not have to make assumptions about the correlation between the baseline and endpoint values like I used in Test2.jpg. In Test2.jpg (rows 2 and 3), I used assumptions of r=0.9 and r=0. These are extreme correlations but it is nice to see they are similar to the SMD calculated using the p value (i.e., that in Test1.jpg). If you want a reference for the CMA software, use the book by Michael Borenstein et al. (i.e., Introduction to Meta-Analysis, Publisher: Wiley [2nd edition], 2021). SD of Difference = SE x square root of 239
Gordon L Warren Hi, thank you for your detailed answer and the attachments. I don't have the baseline and endpoint measurements for each value as it was not provided in the article and I had no luck reaching the authors. all I have is the information I provided. I will read the Michael Bornstein reference, thank you
My language was wrong. I didn't mean you personally had baseline and endpoint measurements for each of the 239 subjects. I meant the study design was such that the authors had baseline and endpoint measurements for each of the 239 subjects. I was uncertain about the statistical test that they used. My assumption was that the original authors used a paired t-test. If so, the effect size calculated in Test1.jpg is correct.
- Badges
- Science topic
- Similar topics
- Mathematics
- Statistics
- Sampling Design
- Sample Size