2
$\begingroup$

I'm looking for a way to conduct a Meta-analysis on Odds Ratio's from several studies/article's. The thing is that my OR's are based on logistic regression's with continuous predictors (outcome is binary, affliction or not). So they are not based on frequencies of hits in experimental en control group. Can I do this analysis with the package metafor in R? And what kind of additional information, such as sample size (n), regression weights and CI's do I need to have R calculate standard errors to estimate the overall effect (OR) and make funnelplots? Hope any one can help me out!

$\endgroup$

1 Answer 1

1
$\begingroup$

You need the coefficient from the logistic regression (in its logged version) with its standard error. You may have to back-calculate it from the confidence interval for the odds ratio which would involve taking logs first. One think to beware of is that if the other variables in the logistic regression are not the same then the coefficients are not estimating the same quantity.

$\endgroup$
1
  • $\begingroup$ Thank you! Makes sense, is your last comment because of unstandardised regression weights or because of individual or partial contribution of age (due to the other covariates? $\endgroup$ Commented Aug 30, 2023 at 11:33

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

Not the answer you're looking for? Browse other questions tagged or ask your own question.