I'm carrying out a meta-analysis on STATA v.13, I'm using the effect size and CI, I transformed them into log scale but some are
How can you use effect size and CI? You have to use standart error (or variance SE^2) with effect size. Another solution is using raw data (crosstable or n + mean + sd). If you don't reach the raw data, you'II calculate the SE(if possible).
Thank you for your comment. I'm following Sterne's recommendations (http://www.biostat.jhsph.edu/~fdominic/teaching/bio656/software/meta.analysis.pdf) of using the effect sizes (and 95%CIs, already provided in the papers) transform them into log scale and then pool them since many reports are lacking of SE is not possible to calculate it. Talking with a colleague he's suggesting to use a constant value to correct values below zero, however I think this is pretty arbitrary. Cheers.
In meta analysis, 1/SE^2 (for fixed effect) or 1/(SE^2+tou^2) (for random effect) is a value that your studies’ weight. If you want to combine OR you ‘II calculate log OR with SE(for log OR). After calculating, multiply the weight by log(or).
I think you should check your operations. If you share your data and your command in stata, I’II deal with it.
I'm carrying out a meta-analysis on STATA v.16, I'm using the two group comparisons of binary outcomes. The result is recorded as "log"OR in forest plot. How could I converse it to OR (only ) in the plot?
- Badges
- Science method
- Similar topics
- Mathematics
- Statistics
- Statistical Techniques
- Meta-Analysis