Читать книгу Introduction to Abnormal Child and Adolescent Psychology - Robert Weis - Страница 133

A final goal of psychological research is replication. We have greater confidence in the results of studies if our findings are reproducible. Replication is especially important in studies that evaluate the efficacy of treatment. We want to be confident that treatment is likely to work before we recommend it to families (Schmidt, 2017).

Not all studies investigating the same phenomena yield identical results. One study may show that therapy greatly improves children’s functioning, another study may indicate that it is only moderately helpful, and a third study may suggest no benefit whatsoever. The studies may also differ in their number of participants and the manner in which they measured children’s outcomes. How can we combine the results of these studies to determine the overall efficacy of therapy and make a decision about its usefulness?

Meta-analysis is a widely used statistical technique to combine the results of multiple research studies into an overall, numerical result (Del Re & Fluckiger, 2018). The result of each study is converted into a single metric called an effect size. The effect size (ES) reflects the magnitude of the difference between the treatment group and the control group at the end of the study. Here is its formula:

ES = Mtreatment group – Mcontrol group

SD

First, we calculate the difference between the mean score of children in the treatment group and the mean score of children in the control group. Then, we divide this difference by the standard deviation of scores (SD), a measure of variability. The result is a single number that reflects how many standard deviations the treatment and control groups are apart. Positive scores indicate that children in the treatment group fared better than children in the control group and therapy was helpful. Negative scores indicate that children in the control group experienced better outcomes than children in the treatment group and therapy was harmful (Hoyt & Del Re, 2018).

We can combine the results of multiple studies by calculating the weighted average effect size. Studies are weighted based on their number of participants, so large studies influence the average more than smaller studies. As a rule of thumb, effect sizes of .2 or less are considered “small,” .5 are “medium,” and .8 or greater are “large” (Ferguson, 2017).