This tutorial will define and discuss meta-analysis including:
Meta-analysis is different than performing a regular experiment. With a regular experiment, someone decides that they're going to perform the experiment, data is collected, then the data is analyzed. That's where the story ends.
But what if someone else does a similar experiment? It may be similar or contradictory. And what if someone else does a third experiment and they get different data, as well?
Sometimes it's useful to pull all those analyses from the different experiments together and put them into a single document. This is called meta-analysis. All the previous work is gathered and synthesized.
Meta-analysis is the process of gathering the data from multiple different experiments that multiple different people have done.
The goal of meta-analysis is not to analyze the data itself, but rather, the overall trends. If the experimental designs were similar, results can be combined to have a more powerful result because of the larger sample size.
Essentially, when this happens, sample sizes from all the previous experiments (with similar designs) are pooled together. This increases the replication, which elicits a powerful result because the overall trends can be seen that may have been missed in smaller experiments. When looking at one experiment, evaluation is limited to the results of only that experiment.
Meta-analysis piggybacks off work that's already been done. The goal is to find overall trends in the data, not to produce raw data. When using meta-analysis, there is opportunity to combine the results of some experiments to find some larger trend and be able to confirm it.
Source: Adapted from Sophia tutorial by Jonathan Osters