Thread by @matherion

Dear #rstats & #systematicreview Twitter friends: I've been working hard on the {metabefor} R package these past weeks. I'll explain a bit more in the thread, but first, my question.

This is the current version of the logo.

(poll with question in thread)

What does this logo evoke?

... And to say a bit more about the package: it's meant to assist with the extraction of data from articles. Not by automating anything, mind you - but by providing a uniform, solid structure to the process and making it transparent, extensible, and machine-readable.

It does so by outlining a standard for specifying exactly what you want to extract from your sources (in a spreadsheet), in a hierarchical structure with the possibility to repeat entities and to refer to other entities, in plain text files that are ultimately R Markdown files.

Those R Markdown files can be immediately rendered, showing the structure of the extracted data, as shown below.

(I admit, this is a crappy example file 😬)

This also immediately shows which values has been extracted, allowing the extractor to check whether everything went well:

(again, this is an example file where I haven't bothered to 'extract' anything but the bare minimum, specifically 'varId' and 'associationId')

It also immediately validates all extracted values according to the validation directives specified (with R expressions) in the original extraction specification referred to above, which, by the way, can be a spreadsheet in Google Sheet or .xlsx format:

Finally, the most recent feature: it can combine extracted data from multiple files. This lends itself well to, e.g., starting with a scoping review and then, in a second phase, extracting more details about some entities to realize a systematic review.

Anyway - enough about this, I'll post a more extensive tweetorial when it's on CRAN. For now, just wanted to share my enthusiasm 🙂