What is the general purpose of meta-analysis, and what are its commonly used procedures?

What will be an ideal response?

Meta-analysis refers to statistical analyses that combine and interpret the results of independent studies of a given scientific issue for the purpose of integrating the findings. The approach treats each study in the review as a case within a sample of studies and applies statistical analysis to each case.

The first thing researchers do is summarize the information abstracted from the studies reviewed based on the coding sheets. Next, the investigator may conduct a more refined subgroup analysis by grouping the studies into comparable categories based on a number of criteria. The results of the subgroup analysis can be compared with the analysis done on the total studies. Such a comparison enables researchers to evaluate whether sources and quality of research are significantly related to differences in results. If they are, investigators can present the meta-analytic results separately for different sources of information and different levels of quality of research. If the findings are mixed, the researcher may conduct a simple vote counting using a sign test to assess whether observed differences are significant.

The most popular and most important meta-analysis procedure is called effect size analysis. Effect size is the size or strength of the impact of one factor on another. In meta-analysis, the effect size is determined by the difference between the control and treatment change scores. Effect size may be estimated by the coefficient of determination.

The first step is to identify the effect size to be investigated, and the second step is the actual computation of an effect size for each study reviewed. In situations where data have already been processed, the effect size can be computed based on the given information. The substantive interpretation of the size of the effect may be made by reference to other information. The most useful interpretation occurs when an effect size is compared with other effect sizes using similar variables.

The standard error measures how accurately the effect size has been measured. The smaller the standard error, the more accurate is the measurement. Once the researcher has calculated effect sizes for all studies, he or she then averages these effects to obtain the mean.

The homogeneity test is used to test for the homogeneity of effect sizes, that is, whether obtained effect sizes were random samples estimating a single effect size or coming from different populations. The result of this test is compared with a chi-square for df = k – 1, where k is the number of studies. If the value is larger than the critical value in the chi-square table, the result becomes significant and suggests the effect sizes did not come from a single population value.