Legend

Source of raw data to be used by studies when evaluating outcomes -- should be non-biased, assuming the data is simply an aggregation of numbers.

Published paper or discussion that has aggregated data from one or more sources, has analyzed said data, has described its methodologies in detail, and has been peer-reviewed.

A synthesis. Ideally a non-biased aggregate of multiple studies. A review considers several published studies, the data therein, compares them against each other, and draws often much larger conclusions based on second-order data models. These are considered the greatest sources of high-level, data-driven truths.