How to Calculate Effect Sizes in Meta-Analysis

A meta-analysis is a study that looks at a large set of individual studies and draws an overall conclusion from their results. In a meta-analysis, one important statistic is the effect size, which essentially is a (scaled) measure of the differences between the two groups of interest, called the treatment group and control group. Calculating the effect size of a meta-analysis requires a researcher to pool together the important statistics from the individual studies before employing calculations.

Instructions

    • 1

      Calculate the treatment mean of the meta-analysis. Locate the treatment means of all of the individual studies. Add these treatment means together, dividing by the number of studies looked at. The resulting number is the meta-analysis treatment mean.

    • 2

      Calculate the control mean of the meta-analysis. Locate the control means of all of the individual studies. Add these treatment means together, dividing by the number of studies looked at. The resulting number is the meta-analysis control mean.

    • 3

      Compute the meta-analysis variance. This variance is the pooled variance of all the studies included in the meta-analysis. Find the sample sizes for each study and label them as "n1" for the first study, "n2 for the second study" and so on. Find the variances for each study and label them "v1" for the first study, "v2" for the second study and so on. Calculate the pooled variance with the equation "Sigma[(ni-1)*(vi)^2]/[Sigma(ni)-g], where "Sigma" represents the function that sums the variables over the entire set of studies and "g" represents the number of studies included in the meta analysis. This pooled variance is the variance for the meta-analysis.

    • 4

      Determine the effect size. Subtract the meta-analysis control mean from the meta-analysis treatment mean. Divide the result by the square of the meta-analysis variance. The resulting number is the effect size for the meta-analysis.

Learnify Hub © www.0685.com All Rights Reserved