Effect Size Conversion and Harmonization of Study Results

Effect size conversion and harmonization are essential steps in meta-analysis because individual studies may report findings using different statistical measures, units of measurement, or reporting formats. This session introduces the purpose of converting effect sizes and aligning effect directions and measurement scales before combining study results. Students will explore the relationships between standardized and unstandardized effect sizes for continuous outcomes, as well as appropriate conversions involving risk ratios (RR), odds ratios (OR), and correlation coefficients. Particular attention will be given to logarithmic transformations of RR and OR and the Fisher (z) transformation of correlation coefficients. These procedures help ensure that effect estimates are expressed on compatible scales and that their direction can be interpreted consistently across studies.

The second component focuses on reconstructing effect estimates and their uncertainty when the necessary statistical information is not reported directly. Students will learn how standard errors and variances can sometimes be derived from confidence intervals, test statistics, p-values, and other available summary statistics. They will also examine approaches for handling incomplete reporting, including situations in which means, standard deviations, sample sizes, or other information must be used to recover an effect estimate. However, such reconstruction depends on the statistical methods, assumptions, and information available in the original study. Students must therefore understand that not every effect measure can be converted reliably into another, and that some conversions require additional assumptions that may introduce uncertainty or information loss.

The final component addresses the challenges of maintaining data consistency and statistical independence before synthesizing results. A single study may report multiple outcomes, several follow-up times, or multiple effect estimates from the same participants, creating dependencies that must be considered during meta-analysis. Students will learn why these estimates should not automatically be treated as independent observations and how inappropriate conversions or inconsistent data handling can distort pooled results and their uncertainty. By the end of this session, students should be able to identify when effect size conversion is appropriate, explain the assumptions and limitations associated with common conversion methods, reconstruct effect estimates from available statistical information when justified, and assess the consistency and dependency of study results before combining them in a meta-analysis.