Avinno Faruk

Very informative #blueskAI 🤖

Thank you!

Out of curiosity, was there any particular reason for prevalence of women as interviewers?

A searchable, linked list of 🏛 #Government agencies at the national, state, and local level who are using GitHub to share and collaborate.

government.github.com/community/

Nice, thanks! 🙏

Haven’t looked into it yet, but did they release the weights of the pre-trained model? A demo I saw only showed a GUI, so just wondering.

Also really helps us students to simply annotate and focus on listening to the lecture, rather than trying to write everything down!

🙋🏻‍♀️

📈📉There’s a kind of paper that uses a “Sherlock Holmes” mode of inference. Start with an interesting fact, lay out the likely suspects, eliminate all but one, and the remaining one is the best explanation. What are some papers that do this well?

Weekly links January 12, 2024: AEA keynotes, therapy and development, Stata vs R popularity over tim... This week's links include AEA keynote talks by Susan Athey and Pol Antras, a review of how psychotherapy helps development outcomes, how to think about climate change and firms, the Stata vs R usage i...

This week's links include insights from AEA keynotes by Susan Athey and Pol Antras, data on the use of Stata vs R in replication packages, an overview of what climate change means for firms in developing countries, therapy and development, and more...

#EconSky this is pretty cool!

Data Imaginist - A small patch of free features patchwork 1.2.0 is out with two outstanding new features. Read on to learn more

Celebrating the re-opening of CRAN with a brand new version of Patchwork 🎉

What can we learn from historical pandemics? A systematic review of the literature by @ainedoran.bsky.social, @cliochris.bsky.social & @eoinaldo.bsky.social.
New and #OA!
doi.org/10.1016/j.so...

🛟 📗📉📈🗃#Episky #history #demography #polisky #geosky #sociology #anthro #HistSci #AcademicSky

Program Evaluation - Program Evaluation for Public Service Combine research design, causal inference, and econometric tools to measure the effects of social programs

Phew, v9.0 of my program evaluation / causal inference class is now live! All the assignments use #QuartoPub and the native |> #rstats pipe, so hopefully nothing catastrophic happens because of all those changes 🤞 evalsp24.classes.andrewheiss.com

Yes I think Stata should consider integrating this feature given how much they charge, lol.

Thanks for sharing!It could be a good workaround. My comments were based on two experiences recently where the user-written package kept getting updated and our results would change every time. version didn’t help as Stata version was not the problem. We ultimately decided to not use those packages.

Thank you for the link! The guide seems to be very helpful. Option 1: I thought that’s only for version control of Stata itself, not for controlling multiple package versions/updates? And option 2 still seems less convenient than, say, generating an automatic requirements file using a Python IDE.

Thanks for the recommendation! This was meant to be a general comment where I used examples from the three I use frequently! And I have been meaning to look into Julia.

Haha no this was meant to be a general note, I just used examples from the three I regularly juggle with :) I had to learn some basic MATLAB for some work early this year but don’t use it regularly.

personal experience over the years, so you’re free to disagree! It’s just I’ve been meaning to write this 🧵 for a while and finally found some time.

End.

(n/n)

main point of my rant was that don’t be a hostage of a tool, instead know your preferences, common applications and best use cases of each, and decide what’s best given a task. Knowing a couple of things simultaneously doesn’t hurt!
Also this is purely from an Econ perspective and based on (10/n)

a quick analysis, it’s easier to write a do file for an econ-related task than a Python script.
The only place I’d draw the line is probably analysis using Excel. It’s still useful to know extremely basic excel of course, e.g. tracking field team activities, budgeting and HR stuff.
I guess the (9/n)

versions, and equally dislike there’s no version control system for non-official ones (at least not that I’m aware of.) Their help files are well documented usually, and I don’t have to decide which is the best package for a specific task like R. Many time if you have smaller datasets and need (8/n)

of variable names, and without having to create separate lists or dictionaries. This is particularly important in domains who frequently share their findings to policymakers and non-academic audiences. I also like the fact that their official commands have backward compatibility across (7/n)

Many economists still continue to release packages in Stata, which are of more value within our discipline only perhaps or to the wider social science community.

What I personally like is the fact that I can use produce graphs and tables very easily using the variable labels directly instead (6/n)

the Econ world has not. E.g., if you’re asking for code from someone about an old paper that you’d like to replicate, more likely than not they’ll just send you a bunch of do files. If you’re involved in past/older projects and joining midway, they still continue to use Stata. (5/n)

It’s not free unlike the other two above. It’s slow when dealing with large datasets. If you’re trying to do ML etc., big no.
With that being said, I still think there’s some value in learning/knowing it. You may have pivoted to the language of your choice, but you should remember the rest of (4/n)

of code in Python is, in short, a pain.
However, I find the R syntax usually annoying and cumbersome for other tasks, especially when I’m trying to write a custom function, for example.

And then, of course, no Econ coding discussion can be complete without mentioning dear old Stata. (3/n)

datasets, the syntax is intuitive, easy to create code documentation and find code samples online for most of your problems.
That being said, when it comes to visualisation, nothing beats R yet among these three. ggplot is beautiful. Getting an aesthetically pleasing visual with minimal lines (2/n)

#EconSky Unpopular opinion, but I really find all this Stata vs R vs Python vs whatnot debates among economists really tiresome.
A 🧵:
Everything has its pros and cons, as well as use cases.
Personally speaking, I gravitate towards Python mostly for most tasks these days. It’s fast for large (1/n)