What Is Effect Size?
Although the statistical calculations are complex, effect size roughly measures the average effect on subjects from several studies of one or more treatments compared with that of a control treatment.
Effect sizes are the meat of meta-analysis, a statistical method developed in 1988 to compare two or more groups that may have little in common to compare. For example: In a hypothetical study comparing the efficacy of risperidone with that of clozapine (the largest actual comparative dataset available to date), researchers might find that the patient populations, assessment methods, or even the doses of the medications used in the different studies under comparison are not the same. As such, direct comparisons across the various studies are difficult.
Using meta-analysis, researchers could determine that all patients taking risperidone, as a group, have an average improvement of so many points on one of the assessment scales, but the distribution of the scores of each individual within the risperidone group covers a range. The same would be true for patients who were treated with clozapine.
The effect size is a measurement of how much the ranges of scores from the two groups overlap.
An effect size of zero would mean the two ranges overlap completely, with their averages being the same. However, a positive effect size for risperidone, for example, would indicate that the average of the risperidone group showed greater improvement than those taking clozapine. A negative effect size would indicate that patients taking risperidone had a lesser degree of improvement than patients in the clozapine group.
An effect size of 0.2 is generally considered “small” and signifies that about 15 percent of patients had greater improvement than the control group (in the example above, clozapine); an effect size of 0.5 is regarded as “medium” and equates to 33 percent of the treatment group having greater improvement compared with the control group; an effect size of 0.8 is regarded as “large” and indicates that about 47 percent of the treatment group had greater improvement compared with the control group.