edgeworth boxlord

So you're making replication files for a simulation paper. You're posting the calibration/rstimation data and the code. Do you post sim output (e.g. CSV with model solutions)?

Would appreciate boosts/replies, I'd like to know what people are doing 🙏

#EconTwitter #rstats #opensci

yes ofc (sim data is data)
52.4%
no (they can run the code)
28.6%
it depends (on what?)
19%
Poll ended at .
Jan 21, 2023 at 1:59pmWeb
Show threadedgeworth boxlordJan 23, 2023
Has to be rounded, but still -- nice 👍
Show threadHenry NgoJan 21, 2023

@akhilrao I say depends because sim data is data, which means it should not be in a code repository. It should be something they can regenerate.

However, if that is time consuming, or if you expect people to just use your sim data as input for additional analysis then the data should be published too. Also if the data is plotted or tabulated in the paper then it should be available as a machine readable format, to help people plot your data in their future work (and cite you!)

Show threadHenry NgoJan 21, 2023

@akhilrao usually the code is on GitHub and the data is elsewhere. If small enough, it could just be in the data supplement files of the paper. Some fields have their own data hosting for this.

Oh also another reason to publish the sim data is if you think the reader might want to validate their reproduction of your code.

Show threadJulian FietkauJan 23, 2023
@akhilrao @ryanstraight In my area of research the question doesn't come up much, but for my PhD thesis I'm including a few small CSV files as PDF attachments to reduce the chance of them getting separated from the manuscript down the road. I'm also attaching the gnuplot scripts which turn those CSV files into my diagrams so anyone can verify those are not fudged in any way. Kinda proud of the whole setup tbh, and it's not something I've had much opportunity to talk about. 😀