It's often hard to get a clear picture of the state-of-the-art research in domains relevant to EA interests. Meta-analyses -- holistic analyses of a number of individual studies -- are stronger evidence about the impact of an intervention than any one study can be by itself, so they are great tools for people trying to use evidence and reason to make the world better. Just as the Cochrane Collaboration exists to evaluate and summarize the best evidence on medicinal practices, effective altruism could benefit from people writing meta-analyses on e.g. iodine fortification, donation-matching programs, or calibration training.
Doing meta-analyses well is hard. We expect this activity to be best suited to people who are comfortable reading and generating research, and doing statistical evaluations.
To get started, check out Luke Muehlhauser's how-to guide on efficient scholarship. Before writing, check out these literature reviews on spaced repetition and factors for job success. Then seek feedback on your post before publishing to the EA Forum, LessWrong, and/or your own blog as makes sense. (h/t Richard Batty)