Dan Scott @dbs@scholar.social

I started my doctoral program in 2018-2019.

No, family, my convocation ceremony has not been cancelled due to COVID-19. Could you at least give me four years before you start asking me about convocation?

Jeez!

Love trying to access a paper behind a paywall, only to get a Cloudfront 504 error.

What was that value that proprietary publishers add, again?

@jon_valdes I love it!

Having gone through the process of learning pandas (I'm a long-time Pythonista) enough to merge, clean, and recode some survey results, I would feel more comfortable loading the data into PostgreSQL and munging the hell out of it there with CTEs, GROUP BYs, and aggregate functions, etc.

But... having a perfectly replicable, automated process from start to finish is nothing to sneeze at.

For the last two weeks our research team has been trying to agree on one number: how many people completed our survey.

I think I've finally convinced them that the number I've given, implemented in Python pandas, is the most correct.

This does not bode well for the real data analysis to follow :/

So just descriptive stats for now. I haven't even started plotting data yet, but I'm looking forward to that.

So much better than fumbling around in a spreadsheet!

I am really liking Python's pandas for cleaning up and analyzing these survey results. I was facing two sets of results for the same survey with responses that were coded differently in each set (due to translation reordering).

So I'm working with the text value of the responses and using https://pandas.pydata.org/pandas-docs/stable/reference/api/pandas.DataFrame.replace.html#pandas.DataFrame.replace to get everything into a single language, then doing a first pass of analysis with https://pandas.pydata.org/pandas-docs/stable/reference/api/pandas.DataFrame.describe.html#pandas.DataFrame.describe . So far so good! And REPLICABLE!

For context, I discussing this with a research team and was surprised they wanted to keep the partial data. Given the preamble, keeping the data seems unethical.

They have since changed their mind. \o/

Consent preamble on a survey that you're taking said "You can choose to withdraw at any time by closing your browser or navigating away."

After you complete about half of the survey, you decide that you don't want to finish it and close the browser.

What do you expect the researchers to do with the partial data you submitted?

I helped design a bilingual survey, but didn't implement it. The PI entered it all into Qualtrics, twice: once for each language.

Now that we have the results, the codes don't match up between languages (of course), so I have to do a ton of cleanup before I can start analysing the data.

But first I have to stop staring at https://www.qualtrics.com/support/survey-platform/survey-module/survey-tools/translate-survey/ wondering why the PI didn't just use that?

@bgcarlisle once everyone taps on the tables you'll know it's all good (and German)

See p.8 of https://doi.org/10.5860/lrts.63n2.119 for my current example. (It's paywalled--stupid ALA--but Google Scholar will find a perfect copy for you at academia.edu)

If I zoom in on-screen, I can barely make out the blurry label text.

Worse, even when printed at 600dpi--theoretically the whole reason for the weird layout--fig. 3 is unreadable.

I hate when I have to struggle to read graph labels in PDF articles on even a 27" QHD screen.

There has to be a better way. Like HTML?

Or at least not arbitrarily scaling the graphic down to 5/8s of the page width. Use the full "8.5"" page-width; it's not like an extra page or two in article length is costing you anything.

@VictorVenema I like that too, but "The person who identified the problem gets to solve the problem" was a principle that I learned at IBM, and it seems to be the default elsewhere...

Today from when you join the review of draft #1 of a collaboratively-edited 30 page whitepaper on how to implement X in an information retrieval system, where X is a cool thing big corporations have been doing for a few years, and ask the question "So has any research been done on whether X actually benefits users? Is there a lit review?", and you get (presumably embarrassed) silence as a response.

Guess who gets to pull together that lit review?

@cassandreces That's a cool project! Can't believe I hadn't heard of it before--thank you (both for sharing it, and for mentoring)

@cassandreces To be honest, I've never tried that; always started with fresh Wikibase instances.

But as Appropedia seems to be a unilingual wiki, maybe Semantic MediaWiki might be a better fit? https://www.semantic-mediawiki.org/wiki/FAQ#What_is_the_relationship_between_Semantic_MediaWiki_and_Wikidata.3F

@cassandreces Right! Barbara Fischer from the German National Library was a participant in the Wikibase workshop Stacy and I led at SWIB18 (https://gitlab.com/denials/wikibase-workshop-swib2018/blob/master/wikibase-workshop-swib18.adoc)