During data extraction for meta analysis, authors sometimes report two subgroups within each treatment arm of the trial. In this condition, we need to include the binary or continuous outcomes of each treatment arm rather than those for each of the subgroups. May I know how can we combine binary or continuous outcomes of two subgroups ?
For binary you can add them up and divide by total.
for continuous data, Cochrane has the following formulae
http://handbook-5-1.cochrane.org/chapter_7/table_7_7_a_formulae_for_combining_groups.htm
Dear Sai Wai
Continious data in Excel:
Mean for two combined groups =
(ObservationsInGroupOne*MeanInGroupOne+ObservationsInGroupTwo*MeanInGroupTwo)/(ObservationsInGroupOne+ObservationsInGroupTwo)
SD for combined groups =SQRT(((ObservationsInGroupOne-1)*SDForGroupOne^2+(ObservationsInGroupTwo-1)*SDForGroupTwo^2+ObservationsInGrouoOne*ObservationsInGroupTwo/(ObservationsInGroupOne+ObservationsInGroupTwo)*(MeanGroupOne^2+MeanGroupTwo^2-2*MeanGroupOne*MeanGroupTwo))/(ObservationsInGroupOne+ObservationsInGroup2-1))
The Campbell Collaborative Web site has a lot of resources that may be useful to you with this issue. Their online calculator is helpful in determining how in translate data into common effect sizes:
https://www.campbellcollaboration.org/escalc/html/EffectSizeCalculator-SMD-main.php
The textbook "Practical Meta-analysis" is also a great resource for people just getting started with meta-analyses. Good luck!
- Badges
- Science method
- Similar topics
- Mathematics
- Statistics
- Statistical Techniques
- Meta-Analysis