IQR changed and removed data#572

It looks like IQR in extraction has changed from intedning to put Q1 and Q3 into one data entry for the range. Is it possible to go back to the old appraoch if desired? additionally, a consequence of this change is that it appears all extracted elements using IQR have had the Q1 and Q3 data deleted. Can this be restored?

Thanks for the heads up Derek! This was changed with release 1.99.0; I’ll look into it and get back to you by this afternoon with a solution.

This was removed to make way for Smart MA Extraction in release 1.99.0. Smart MA Extraction (currently) is somewhat opinionated in how it extracts data in the name of generality. WRT median/IQR, it prefers IQR as the true statistic definition (the difference of the 25th and 75th percentiles, not the percentiles themselves).

The new approach generalizes better because if the source paper reports percentiles, the IQR is easily derived as the difference. If the IQR is instead collected as the [25th, 75th] percentiles and the source paper reports only the IQR, there’s no way to properly infer the range values.

That was the intent of the new development, which I think was well reasoned… However, we cannot break backwards compatibility. We’ll roll that change back (old data will be left in place as it was before) and figure out a separate change to accomodate Smart MA Extraction’s approach.

Also, I missed an apology for the breakage & scare of missing data– sorry from the dev team!

Thanks for the response and for working on the solution! This all makes a lot of sense and I look forward to the Smart Extraction roll out as well!

Thanks Karl! We do see some studies with the true definition of IQR, and it would be great to have the option to grab those, but the majority report a range. For continued ease of the human extraction option and easy quality checking, I’d vote for keeping an option to extract IQR as a range if possible. Looking forward to trying out the Smart MA extraction though!

Thanks!

This should be corrected as of release 1.99.3, albeit with a relabel to “Q1-Q3”. Note that when configuring new DEs dispersion = “IQR” will correspond to the singular IQR statistic. Please let me know if you have any questions– thanks!

Thanks Karl. This is a great solution and is working on my end.

seconding Derek, all of the previous data is restored and this works great for us to have the option to do either! Thanks!

Awesome to hear, thanks for confirming!