Abstract:
|
It is not uncommon for a researcher to dichotomize a continuous outcome for analysis. If the study has two treatment arms, the effect size of interest could change from a standardized mean difference to a relative risk. This is problematic in a meta-analysis where study effect sizes and their variance estimates must be on the same scale. To alleviate this problem, we propose methods for transforming the relative risk and odds ratio to a standardized mean difference in terms of the probability of treatment arm allocation. Variance estimates are also derived. The results from our methods are compared with previous transformations of the odds ratio to the standardized mean difference proposed by Chinn (2000) and Chen et al (2010).
|