For a meta-analysis, I'm including some studies that report the exposure variable in hours as categorical variables (e.g., "1 to 60 mins"; "61 mins to 120 mins"; "121 mins to 180 mins", etc.), with each category compared to the same referent (e.g., 0 hrs of exposure) and an odds ratio reported for each. While I'm aware of how to compute an effect size based on odds ratios from 2 x 2 data (e.g., exposure yes/no & outcome is present/absent), I can't find info on how to compute effect sizes from odds ratios where there are multiple levels of exposure for an outcome.
Is there a way to capture one effect size based that incorporates all of the categories, or do I need to compute multiple effect sizes for each as distinct data / "studies", although each uses the same referent group? Any suggestions or references would be greatly appreciated.
If there is sufficient raw data in the published studies or you can contact original authors, it may be possible to calculate logistic regression coefficients for dose response assuming continuous independent variables and then estimate a meta coefficient.
See eg Rothman and Greenland ,Modern Epidemiology.
- Badges
- Science topic
- Similar topics
- Engineering
- Computer Engineering
- Computer