Meta-analysis: Statistical alchemy for the 21st century

(1) The substances under investigation should be identified in a reproducible manner. (2) Each substance should receive a precise characterization, which indicates its specific distinctiveness. To achieve this goal, scientists search for substances that have been suitably purified, and try to work with mixtures that are homogeneous, not heterogeneous. (3) Comparisons that are done within a study should be fair or unbiased. In a randomized clinical trial, this attribute is called the internal validity of the comparison. (4) What is learned in the research should be suitable for extrapolation and application elsewhere beyond the restrictions of the immediate setting and focus of the research. In randomized trials, this attribute is called external validity.