A meta-analysis is an analytical review of several research studies on a common topic. Scientific research is based on statistical findings, but these studies are often limited by sample size, as only a small sample of possible data can be gathered in the course of any given project. Meta-analysis aims to overcome this difficulty by combining the findings from several studies, creating a more comprehensive picture of the research problem. Although this type of analysis has advantages, it also has drawbacks, such as selection bias and possible distortion of statistics that may lead to false conclusions.
Meta-analyses can be done in any area of study where a body of statistical research literature exists. In order for the analysis to be valid, however, it must be done systematically, like a research study itself. After the problem is formulated, certain studies are chosen for inclusion in the analysis based on specific criteria.
The nature of the criteria depends on the goal of the meta-analysis. For example, a researcher conducting a meta-analysis on treatments for patients suffering from heart attacks would include studies specifically about this topic. The researcher might further narrow the selection of literature by choosing only studies that were performed with an appropriate methodology. For instance, the requirement for randomization, or random selection of samples to prevent bias, might be a criterion for inclusion.
After the studies have been collected and reviewed, statistical methods are used to combine and filter data. Since the sample size in a meta-analysis is effectively much larger than the sample size in an ordinary research study, it may be possible for the analysis to reveal statistical patterns that a single study could not show. The small sample size of a single research study can also magnify certain chance effects out of proportion. Meta-analysis can be used to resolve contradictions between studies arising from such random fluctuations.
Perhaps the greatest drawback of the meta-analysis process is the problem of selection. Since the researcher must choose which studies to include in his or her analysis, a bias in the overall statistical conclusions is unavoidable. A researcher with a given agenda could conceivably skew the selection to favor certain conclusions over others. Even if the topic of the analysis is narrow enough that all available literature can be reviewed, unpublished studies will not be included. Critics of meta-analysis point to this as evidence that the process is not truly objective or scientific.