How Can I Learn to Write Meta-Analyses?

Meta-analyses are first and foremost about deriving meaning from the data across many papers on the same topic. To do them, you'll need a combination of literature review, data extraction, analysis, and synthesis of the information to derive meaning. There are many comprehensive courses on websites like Skillshare that can dive deeply into the methodology. They have a free trial, too. For this guide, we'll walk through the general stages of a meta-analysis and how to get better at doing them.

A meta-analysis methodology workflow looks like: literature search, screening, store, read, extract, synthesize.

If you've completed literature reviews in the past, you'll know the basics of literature search. Compile peer-reviewed literature from an academic publisher database. Ensure that the methods are reproducible by other researchers by starting with a boolean search. The screening step is when you determine which articles you're going to include and exclude from full-text analysis. In some meta-analyses you're starting with thousands of articles and you'll need to filter these articles quickly to find the articles most valuable to answering your research question. To learn how to create reproducible searches and screening, read "Writing a Reproducible and Replicable Literature Review".

The next step is storage. You'll want to have a predefined grid (an Excel sheet works fine) with the data you plan to extract from articles. You'll populate this sheet quickly with table data and full-text keyword search. An example table looks like this:

You populate this table using data extracted from the research articles. You can extract data from articles by making forms that you'll fill out for each article as you go through them one by one to make the task simpler, or by adding directly to a spreadsheet. In some fields, research papers will have descriptive statistics and results for different populations in tables. It's important that whatever you extract is in a format that can easily be used within the statistical software that you use for the analysis portion.

Note: Some researchers who do meta-analysis often will extract tables and save them before the screening process so they are available for future reviews.

A reference manager like Zotero will keep the citation information ready for later. Then you'll want to have an extraction tool like FoundThat ready for pulling out the information on your deep-dive. The simpler the tool, the better. If you plan to use AI with your literature review, check your University policy as some schools don't allow loading papers into an LLM. Keep in mind that AI tools will use methods that may not be reproducible or have bias toward some articles. It's safest to use non-AI tools and spare yourself the headache.

Once you have your literature review portion concluded and you're ready to analyze your data, add your extracted data into your preferred cleaning/transforming software. This can be as simple as an Excel or Google Sheets document. You'll want to have something you can come back to and easily verify your data later. This also helps to check and make sure you've not added the wrong data to analysis.

For analysis, you'll then load your document into a statistical analysis software like R Studio (free) or SPSS (paid). Some colleges and universities will offer SPSS for free to students. It's not the easiest to learn, but once you get past the learning curve, it's a life saver.

If you have more questions on how to keep the motivation through the process, check out "How to Write a Literature Review Without Going Crazy".