Last weekend I attended the Open Economics sprint about Metametrik, which is an OKF project to help computers read the statistics in academic journal articles. Guo Xu posted a rough outline of what Metametrik should look like a year ago, but not much had been implemented.
We took an approach which turned out to be fairly similar to Guo’s recommendation: create a JSON schema capturing the basics of a regression result (the dependent variable, the goodness of fit, the sample size, standard errors and effect sizes of results), and then make tools which produce and consume data in this format.
So far, we have a tool which generates Metametrik format data, and a tool which reads this into a database. What’s needed next is web UIs that produce this data (for articles already published) and allow you to search it.
If you could bear reading this far, you should follow me on Twitter: