I am wondering if we can use the coefficients of Cox regression and logistic regression as a correlation coefficient effect size in meta- analysis if the coefficients reported in a primary study resides in the interval ±0.5(Peterson and Brown,2005). I know that it is still controversial whether to integrate beta coefficients from regression model into analysis where effect size metric is correlation coefficient, but I am worried that omitting these relevant effect sizes would increase my sampling error, which decreases the precision of meta-analytic estimates of population parameters.
how can I solve this problem?
Have a nice day
I do not have a solution for your request, but I see another maybe relevant issue. There is no consensus how to standardize coefficients of a logistic regression. There are several approaches, but please see the following articles for some recommendations:
Menard, S. (2004). Six Approaches to Calculating Standardized Logistic Regression Coefficients, The American Statistician, 58 (3).
Menard, S. (2011). Standards for Standardized Logistic Regression Coefficients. Social Forces, 89(4).
You might find some useful suggestions here:
- https://www.meta-analysis.com/downloads/Meta-analysis%20Converting%20among%20effect%20sizes.pdf
And here:
- https://pdfs.semanticscholar.org/1d95/1c8c87e41d4e0821cb075d7f8cab237bc6c3.pdf
Thank you so much for the paper Rainer Duesing , I hope I will handle the problem.
- Badges
- Science topic
- Similar topics
- Statistics
- Survey