preview

What Are The Different Weighting Methods Used For Combining Effect Size In Studies When Conducting Meta-Analysis

Good Essays

Results
This section describes some of the different weighting methods for combining effect sizes across studies when conducting meta-analysis. This section will focus on methods used for binary outcome data and effect measures such as relative risks and odds ratios. The section will first introduce fixed effect and random effects analyses. It will then describe weighting methods that can be used for these analytic approaches.

Fixed effect versus random effects analyses
Fixed effect and random effects models are the two most common approaches used when conducting a meta-analysis (Borenstein, Hedges, Higgins, & Rothstein, 2010). The distinction between these approaches is both conceptual and statistical in nature. The fixed-effect model …show more content…

Although both fixed-effect and random-effects models will generally give more weight to larger studies with lower variance estimates, the weights will be relatively more balanced with a random-effects model because it assumes that each of the effect sizes in the analysis are estimating a different ‘true’ effect (Borenstein, Hedges, & Rothstein, 2007). Thus, random-effects models will give more weight to smaller studies with higher variance estimates compared to fixed-effect models.

In contrast, the weights for a fixed-effect model will be more extreme than those of a random-effects model because it is assumed that the ‘best’ estimate of the effect size will be from the largest study with the lowest variance (Borenstein et al., 2007). Consequently, compared to a random-effects model, the pooled or summary effect for a fixed-effect model will be closer to the effect obtained from the largest study with the lowest variance. With respect to the variance estimates of the pooled or summary effect, random-effects model will result in higher variances than fixed-effect models because of the added between-study variability assumed to be present.

The assumptions in a fixed-effect analysis may only be appropriate in a limited number of situations. For example, if studies included in a meta-analysis were conducted using a similar methodology by recruiting participants in a similar manner and

Get Access