Elliott Collins
@elliottcollins.bsky.social
Development economics, statistics and philosophy of poverty measurement. Recently returned to Berkeley and Social Media.
elliottcollins.com/
P.S. - All my re-posts are, without exception, endorsements.
elliottcollins.com/
P.S. - All my re-posts are, without exception, endorsements.
Really does seem like a shame. It would be great to hear more from the ones leading/managing the network.
November 17, 2025 at 7:03 AM
Really does seem like a shame. It would be great to hear more from the ones leading/managing the network.
Makes sense. Especially troubling since it's likely averaging across some unaffected and some heavily affected students within cohorts.
November 17, 2025 at 1:34 AM
Makes sense. Especially troubling since it's likely averaging across some unaffected and some heavily affected students within cohorts.
Wait, what's vent?
November 16, 2025 at 6:40 PM
Wait, what's vent?
I did look and the study has two definitions, but I didn't fully understand the one in the main specification. Maybe I was just feeling slow that day.
November 16, 2025 at 4:33 PM
I did look and the study has two definitions, but I didn't fully understand the one in the main specification. Maybe I was just feeling slow that day.
Ken Burns documentaries are available and affordable, to be fair.
November 16, 2025 at 7:08 AM
Ken Burns documentaries are available and affordable, to be fair.
Is this just how the field gets better methods over time or a sign that we're not actually making much progress at all?
November 14, 2025 at 9:19 PM
Is this just how the field gets better methods over time or a sign that we're not actually making much progress at all?
Agreed. As with OS's and text editors, I've cycled through nearly every setup possible at this point and all the big options can do all the stuff you need and they're all perfectly fine.
November 14, 2025 at 8:28 AM
Agreed. As with OS's and text editors, I've cycled through nearly every setup possible at this point and all the big options can do all the stuff you need and they're all perfectly fine.
This is why R makes a lot of sense in an academic research environment where you have a lot of flexibility over the tools you use and progressively less as you get into applied/industry contexts where DS has to integrate seamlessly with non-statistical code bases, which is where DS was born.
November 14, 2025 at 8:02 AM
This is why R makes a lot of sense in an academic research environment where you have a lot of flexibility over the tools you use and progressively less as you get into applied/industry contexts where DS has to integrate seamlessly with non-statistical code bases, which is where DS was born.
It's a fine point for his context, but his article actually starts with a bunch of practical considerations that (perhaps unintentionally) make Python seem like the better choice for most DS students.
November 14, 2025 at 7:57 AM
It's a fine point for his context, but his article actually starts with a bunch of practical considerations that (perhaps unintentionally) make Python seem like the better choice for most DS students.
I feel like people are being more than 40% harder on 50-year loans than they are on 30-year loans.
November 13, 2025 at 11:23 PM
I feel like people are being more than 40% harder on 50-year loans than they are on 30-year loans.
Going to share this with my middle schooler tonight before math homework.
November 13, 2025 at 10:14 PM
Going to share this with my middle schooler tonight before math homework.
"Left viewed" is an interesting phrase. What are you hoping your followers do beyond viewing your posts?
November 13, 2025 at 9:47 PM
"Left viewed" is an interesting phrase. What are you hoping your followers do beyond viewing your posts?
He actually does base R in a footnote. It's a one-liner in base R, but he defaults to the tidyr approach which sort of undercuts his "simpler is better" point.
November 13, 2025 at 9:33 PM
He actually does base R in a footnote. It's a one-liner in base R, but he defaults to the tidyr approach which sort of undercuts his "simpler is better" point.
Does that not make it better? It seems like it does.
November 13, 2025 at 9:22 PM
Does that not make it better? It seems like it does.
That's also why "using no special data wrangling package" there at the end is a little silly. Take the drill bit out of a drill and suddenly it can barely make holes at all.
November 13, 2025 at 9:20 PM
That's also why "using no special data wrangling package" there at the end is a little silly. Take the drill bit out of a drill and suddenly it can barely make holes at all.
An academic research environment (especially where you're in charge) naturally misses the point of using Python.
DS defaults to it *because* it's a general-purpose language, not purpose-built for stats. The field developed from the need integrate data analysis into coding & engineering workflows.
DS defaults to it *because* it's a general-purpose language, not purpose-built for stats. The field developed from the need integrate data analysis into coding & engineering workflows.
November 13, 2025 at 9:14 PM
An academic research environment (especially where you're in charge) naturally misses the point of using Python.
DS defaults to it *because* it's a general-purpose language, not purpose-built for stats. The field developed from the need integrate data analysis into coding & engineering workflows.
DS defaults to it *because* it's a general-purpose language, not purpose-built for stats. The field developed from the need integrate data analysis into coding & engineering workflows.