Thom Volker
@thomvolker.bsky.social
PhD Candidate in Statistics, Utrecht University
Creates fake data for a living.
thomvolker.github.io
Creates fake data for a living.
thomvolker.github.io
This is getting out of hand
November 25, 2025 at 11:23 AM
This is getting out of hand
Reposted by Thom Volker
I wish I didn’t have to share this. But the BBC has decided to censor my first Reith Lecture.
They deleted the line in which I describe Donald Trump as “the most openly corrupt president in American history.” /1
They deleted the line in which I describe Donald Trump as “the most openly corrupt president in American history.” /1
November 25, 2025 at 9:26 AM
I wish I didn’t have to share this. But the BBC has decided to censor my first Reith Lecture.
They deleted the line in which I describe Donald Trump as “the most openly corrupt president in American history.” /1
They deleted the line in which I describe Donald Trump as “the most openly corrupt president in American history.” /1
Reposted by Thom Volker
Aan onze nieuwsredacties:
November 25, 2025 at 9:39 AM
Aan onze nieuwsredacties:
Seems like an awesome project!
CANSSI postdoc w/ Alex Stringer (Waterloo) www.alexstringer.ca and me (McMaster): implementing/exploring Gauss-Hermite quadrature methods in lme4/glmmTMB canssi.ca/wp-content/u... (Alex has shown that Laplace approx is sometimes awful, we'd like to provide alternatives!) canssi.ca/program/dist...
Distinguished Postdoctoral Fellowships – CANSSI
canssi.ca
November 24, 2025 at 10:17 PM
Seems like an awesome project!
Reposted by Thom Volker
Wierd Duk heeft groot gelijk. Ik zou ook opstaan en weglopen als iemand naast mij ineens allemaal teksten gaat zitten optrommelen van Wierd Duk
November 23, 2025 at 9:31 AM
Wierd Duk heeft groot gelijk. Ik zou ook opstaan en weglopen als iemand naast mij ineens allemaal teksten gaat zitten optrommelen van Wierd Duk
Wake me up when we octuple robust estimators
New paper: Triply Robust Panel Estimators (TROP) by @susanathey.bsky.social @guidoimbens.bsky.social Zhaonan Qu @vivianodavide.bsky.social.
arxiv.org/pdf/2508.21536
arxiv.org/pdf/2508.21536
November 24, 2025 at 12:27 AM
Wake me up when we octuple robust estimators
Reposted by Thom Volker
The 28 points would strip Ukraine of its sovereignty and security. It would reward Putin for his aggression and thereby encourage him to go even further. This is unacceptable.
November 23, 2025 at 6:45 PM
The 28 points would strip Ukraine of its sovereignty and security. It would reward Putin for his aggression and thereby encourage him to go even further. This is unacceptable.
Reposted by Thom Volker
I wrote something about z-statistics and the signal-to-noise ratio.
November 23, 2025 at 7:23 PM
I wrote something about z-statistics and the signal-to-noise ratio.
Reposted by Thom Volker
An upd report on our workshops for Ukraine series on #RStats, #Python & more. So far, we have raised >113k euro & the workshops have been attended by > 5000 ppl!
👇you can find more info on how you can help & a detailed report.
All info on workshops: bit.ly/3wBeY4S 1/n
👇you can find more info on how you can help & a detailed report.
All info on workshops: bit.ly/3wBeY4S 1/n
November 23, 2025 at 10:01 AM
An upd report on our workshops for Ukraine series on #RStats, #Python & more. So far, we have raised >113k euro & the workshops have been attended by > 5000 ppl!
👇you can find more info on how you can help & a detailed report.
All info on workshops: bit.ly/3wBeY4S 1/n
👇you can find more info on how you can help & a detailed report.
All info on workshops: bit.ly/3wBeY4S 1/n
Reposted by Thom Volker
Finally understood TMLE’s “doubly robust” property through simulation. Works well when either outcome OR treatment model is correct. XGBoost + TMLE captured complex relationships without manual specification. It worked on simulated complex data, would it work in real world? 🤔 #rstats #idsky #episky
Bias, Variance, and Doubly Robust Estimation: Testing The Promise of TMLE in Simulated Data | Everyday Is A School Day
Finally understood TMLE's "doubly robust" property through simulation. Works well when either outcome OR treatment model is correct. XGBoost + TMLE captured complex relationships without manual specif...
www.kenkoonwong.com
November 17, 2025 at 1:55 AM
Reposted by Thom Volker
But if you have N=100,000 (not at all unlikely with NHANES/biobank-type data), absolutely any noise will be significant at 0.05, so it's a ludicrously large value.
November 22, 2025 at 11:52 PM
But if you have N=100,000 (not at all unlikely with NHANES/biobank-type data), absolutely any noise will be significant at 0.05, so it's a ludicrously large value.
Reposted by Thom Volker
I have been feeling depressed and discouraged that this man has the power and influence he does at my school and in my profession
And I am a tenured professor at Harvard! How much more protected can I be?
Imagine how STUDENTS feel. Junior faculty. This quote nails it
And I am a tenured professor at Harvard! How much more protected can I be?
Imagine how STUDENTS feel. Junior faculty. This quote nails it
November 19, 2025 at 6:27 PM
I have been feeling depressed and discouraged that this man has the power and influence he does at my school and in my profession
And I am a tenured professor at Harvard! How much more protected can I be?
Imagine how STUDENTS feel. Junior faculty. This quote nails it
And I am a tenured professor at Harvard! How much more protected can I be?
Imagine how STUDENTS feel. Junior faculty. This quote nails it
Reposted by Thom Volker
Friendly reminder that a few sponsored registrations are still available. Priority given to black, indigenous people of color in high-income countries or those from low and middle-income countries.
Did I tweak nearly every slide of my regression deck in preparation for next month’s charity course? Only one way to find out!
Join us by dropping a 50 USD donation to World Central Kitchen or United Farm Workers. A few sponsored spots available. Details at betanalpha.github.io/courses/.
Join us by dropping a 50 USD donation to World Central Kitchen or United Farm Workers. A few sponsored spots available. Details at betanalpha.github.io/courses/.
November 19, 2025 at 5:38 AM
Friendly reminder that a few sponsored registrations are still available. Priority given to black, indigenous people of color in high-income countries or those from low and middle-income countries.
Reposted by Thom Volker
Reposted by Thom Volker
November 19, 2025 at 11:41 AM
Reposted by Thom Volker
Few things bug me more than higher ed leaders saying that we lost our mission and lost the trust of the public, when we have actually been the target of a decades-long smear campaign by the right wing that worked. The moment we’re losing our mission is right now, in capitulation.
November 19, 2025 at 1:03 PM
Few things bug me more than higher ed leaders saying that we lost our mission and lost the trust of the public, when we have actually been the target of a decades-long smear campaign by the right wing that worked. The moment we’re losing our mission is right now, in capitulation.
One of my students today: "Ugh next week on Tuesday I have statistics classes from 9 to 5, I'd rather die"
Me doing statistics every weekday from 9 to 5:
Me doing statistics every weekday from 9 to 5:
a cartoon of homer simpson standing in a grassy area
ALT: a cartoon of homer simpson standing in a grassy area
media.tenor.com
November 19, 2025 at 2:18 PM
One of my students today: "Ugh next week on Tuesday I have statistics classes from 9 to 5, I'd rather die"
Me doing statistics every weekday from 9 to 5:
Me doing statistics every weekday from 9 to 5:
Reposted by Thom Volker
Did I tweak nearly every slide of my regression deck in preparation for next month’s charity course? Only one way to find out!
Join us by dropping a 50 USD donation to World Central Kitchen or United Farm Workers. A few sponsored spots available. Details at betanalpha.github.io/courses/.
Join us by dropping a 50 USD donation to World Central Kitchen or United Farm Workers. A few sponsored spots available. Details at betanalpha.github.io/courses/.
November 18, 2025 at 5:26 AM
Did I tweak nearly every slide of my regression deck in preparation for next month’s charity course? Only one way to find out!
Join us by dropping a 50 USD donation to World Central Kitchen or United Farm Workers. A few sponsored spots available. Details at betanalpha.github.io/courses/.
Join us by dropping a 50 USD donation to World Central Kitchen or United Farm Workers. A few sponsored spots available. Details at betanalpha.github.io/courses/.
Reposted by Thom Volker
Enorme heisa om privébericht Wijers, intussen gaat PVV’er met haatpagina weer fluitend aan het werk
Enorme heisa om privébericht Wijers, intussen gaat PVV’er met haatpagina fluitend aan het werk
www.volkskrant.nl
November 17, 2025 at 6:41 PM
Enorme heisa om privébericht Wijers, intussen gaat PVV’er met haatpagina weer fluitend aan het werk
Reposted by Thom Volker
hot take: if people think they can only fit straight lines with linear regression, they should stay yards away from something more complex than linear regression
Confusing stats language:
The "linear" is linear regression means "linear in the parameters".
It does not mean "can only fit straight lines", and things like polynomials and splines can be included in linear models.
online.stat.psu.edu/stat501/less...
1/2
The "linear" is linear regression means "linear in the parameters".
It does not mean "can only fit straight lines", and things like polynomials and splines can be included in linear models.
online.stat.psu.edu/stat501/less...
1/2
November 14, 2025 at 10:00 AM
hot take: if people think they can only fit straight lines with linear regression, they should stay yards away from something more complex than linear regression
Reposted by Thom Volker
lets gooooo! Do some math with me
❗️Our next workshop will be on Dec 11 6 pm CET titled A Gentle Introduction to Mathematical Simulation in R by
@damiepak.bsky.social
Register or sponsor a student by donating to support Ukraine!
Details: bit.ly/3wBeY4S
Please share!
#AcademicSky #EconSky #RStats
@damiepak.bsky.social
Register or sponsor a student by donating to support Ukraine!
Details: bit.ly/3wBeY4S
Please share!
#AcademicSky #EconSky #RStats
November 14, 2025 at 11:17 AM
lets gooooo! Do some math with me
hot take: if people think they can only fit straight lines with linear regression, they should stay yards away from something more complex than linear regression
Confusing stats language:
The "linear" is linear regression means "linear in the parameters".
It does not mean "can only fit straight lines", and things like polynomials and splines can be included in linear models.
online.stat.psu.edu/stat501/less...
1/2
The "linear" is linear regression means "linear in the parameters".
It does not mean "can only fit straight lines", and things like polynomials and splines can be included in linear models.
online.stat.psu.edu/stat501/less...
1/2
November 14, 2025 at 10:00 AM
hot take: if people think they can only fit straight lines with linear regression, they should stay yards away from something more complex than linear regression
Research funding organizations would get so much more bang for their buck if they would collectively bargain a better deal with major publishers, or simply forbid the use of grant money for publishing fees
November 14, 2025 at 9:30 AM
Research funding organizations would get so much more bang for their buck if they would collectively bargain a better deal with major publishers, or simply forbid the use of grant money for publishing fees
Reposted by Thom Volker
We didn't plagiarize, you made us plagiarize by asking questions to which we stole the answers.
"Because its output is generated by users of the chatbot via their prompts, OpenAI said, they were the ones who should be held legally liable for it – an argument rejected by the court."
"Because its output is generated by users of the chatbot via their prompts, OpenAI said, they were the ones who should be held legally liable for it – an argument rejected by the court."
ChatGPT violated copyright law by ‘learning’ from song lyrics, German court rules
OpenAI ordered to pay undisclosed damages for training its language models on artists’ work without permission
www.theguardian.com
November 14, 2025 at 2:47 AM
We didn't plagiarize, you made us plagiarize by asking questions to which we stole the answers.
"Because its output is generated by users of the chatbot via their prompts, OpenAI said, they were the ones who should be held legally liable for it – an argument rejected by the court."
"Because its output is generated by users of the chatbot via their prompts, OpenAI said, they were the ones who should be held legally liable for it – an argument rejected by the court."