Angus Hughes
@awhug.bsky.social
data + statistics enthusiast and one-time psychology researcher
Wattamolla! My best mate and I tried camping out on the beach there under the stars about a decade ago. Nothing but our sleeping bags and a fire. All very adventurous and romantic until we got rained out at 1am. Stunning part of the world though.
October 15, 2025 at 2:41 AM
Wattamolla! My best mate and I tried camping out on the beach there under the stars about a decade ago. Nothing but our sleeping bags and a fire. All very adventurous and romantic until we got rained out at 1am. Stunning part of the world though.
Taking this further, my 3 year old daughter scarcely has any concept of physical money. She vaguely understands the idea of coins because of the claw machine episode of Bluey. Cash I’d wager no clue.
You can get by in Australia for 99% of transactions with a card, or even just your mobile.
You can get by in Australia for 99% of transactions with a card, or even just your mobile.
September 9, 2025 at 10:56 PM
Taking this further, my 3 year old daughter scarcely has any concept of physical money. She vaguely understands the idea of coins because of the claw machine episode of Bluey. Cash I’d wager no clue.
You can get by in Australia for 99% of transactions with a card, or even just your mobile.
You can get by in Australia for 99% of transactions with a card, or even just your mobile.
Reposted by Angus Hughes
The same man whose eulogy for his dad in the Guardian last year made me ugly cry and call my own dad to tell him I love him.
Give him whatever he wants.
Give him whatever he wants.
August 6, 2025 at 9:14 PM
The same man whose eulogy for his dad in the Guardian last year made me ugly cry and call my own dad to tell him I love him.
Give him whatever he wants.
Give him whatever he wants.
You could get the best of both worlds with a raincloud plot
August 5, 2025 at 1:53 AM
You could get the best of both worlds with a raincloud plot
I don’t think mandating social sciences or humanities solves this unfortunately.
I kind of think there are plenty of people who pursue STEM as superior and are clever enough to infer the ‘correct’ signalling or language required pass classes or workplace training, for example, but do so cynically.
I kind of think there are plenty of people who pursue STEM as superior and are clever enough to infer the ‘correct’ signalling or language required pass classes or workplace training, for example, but do so cynically.
August 2, 2025 at 11:55 AM
I don’t think mandating social sciences or humanities solves this unfortunately.
I kind of think there are plenty of people who pursue STEM as superior and are clever enough to infer the ‘correct’ signalling or language required pass classes or workplace training, for example, but do so cynically.
I kind of think there are plenty of people who pursue STEM as superior and are clever enough to infer the ‘correct’ signalling or language required pass classes or workplace training, for example, but do so cynically.
So linking back to the original point, STEM itself isn’t the problem, it’s the unwillingness of some to really engage with ‘lesser’ humanities and social science topics in the first place.
You can easily find these STEM types dismiss introductory ideas in pol sci, feminist studies etc out of hand.
You can easily find these STEM types dismiss introductory ideas in pol sci, feminist studies etc out of hand.
August 2, 2025 at 11:40 AM
So linking back to the original point, STEM itself isn’t the problem, it’s the unwillingness of some to really engage with ‘lesser’ humanities and social science topics in the first place.
You can easily find these STEM types dismiss introductory ideas in pol sci, feminist studies etc out of hand.
You can easily find these STEM types dismiss introductory ideas in pol sci, feminist studies etc out of hand.
Yeah I’m inclined to agree (from an Australian perspective).
The unfortunate result I think though is an attitude of wilful ignorance or dismissiveness to certain ideas in the humanities and social sciences amongst more STEM-oriented folk, and viewing the fields as unserious and prone to sophistry.
The unfortunate result I think though is an attitude of wilful ignorance or dismissiveness to certain ideas in the humanities and social sciences amongst more STEM-oriented folk, and viewing the fields as unserious and prone to sophistry.
August 2, 2025 at 10:56 AM
Yeah I’m inclined to agree (from an Australian perspective).
The unfortunate result I think though is an attitude of wilful ignorance or dismissiveness to certain ideas in the humanities and social sciences amongst more STEM-oriented folk, and viewing the fields as unserious and prone to sophistry.
The unfortunate result I think though is an attitude of wilful ignorance or dismissiveness to certain ideas in the humanities and social sciences amongst more STEM-oriented folk, and viewing the fields as unserious and prone to sophistry.
Call it STEM as Smarter Than Every (other) Major
August 2, 2025 at 7:39 AM
Call it STEM as Smarter Than Every (other) Major
The inverse to this take that I commonly see online is that STEM students could so easily and naturally intellectually dominate a humanities and/or social science curriculum that these topics are essentially beneath them.
I’ve sometimes wondered if these arguments exist to counter one another.
I’ve sometimes wondered if these arguments exist to counter one another.
August 2, 2025 at 7:38 AM
The inverse to this take that I commonly see online is that STEM students could so easily and naturally intellectually dominate a humanities and/or social science curriculum that these topics are essentially beneath them.
I’ve sometimes wondered if these arguments exist to counter one another.
I’ve sometimes wondered if these arguments exist to counter one another.
@andytimm.bsky.social has written a very good blog about how the psychological motivation for scaling best and worst options jointly as MaxDiff does is essentially nonsensical as well, and leads to poorer performance
andytimm.github.io/posts/doing_...
andytimm.github.io/posts/doing_...
Better discrete choice modeling through the rank ordered logit – Andy Timm
Or- a mathematically correct model of a psychologically coherent concept
andytimm.github.io
July 29, 2025 at 4:52 AM
@andytimm.bsky.social has written a very good blog about how the psychological motivation for scaling best and worst options jointly as MaxDiff does is essentially nonsensical as well, and leads to poorer performance
andytimm.github.io/posts/doing_...
andytimm.github.io/posts/doing_...
Average their predictions and, like it or not, the machine is learning, baby
July 17, 2025 at 2:33 AM
Average their predictions and, like it or not, the machine is learning, baby
Agree they’re the worst, and exploded logit is probably the most convenient.
You can fit it as a cox regression, so any software for that should work (see statisticalhorizons.com/wp-content/u...)
I tried this with some NBA draft ranks a few years ago here: gist.github.com/awhug/46d5d3...
You can fit it as a cox regression, so any software for that should work (see statisticalhorizons.com/wp-content/u...)
I tried this with some NBA draft ranks a few years ago here: gist.github.com/awhug/46d5d3...
Mock NBA Draft Order using Rank Order Logit Model
Mock NBA Draft Order using Rank Order Logit Model. GitHub Gist: instantly share code, notes, and snippets.
gist.github.com
January 7, 2025 at 7:07 AM
Agree they’re the worst, and exploded logit is probably the most convenient.
You can fit it as a cox regression, so any software for that should work (see statisticalhorizons.com/wp-content/u...)
I tried this with some NBA draft ranks a few years ago here: gist.github.com/awhug/46d5d3...
You can fit it as a cox regression, so any software for that should work (see statisticalhorizons.com/wp-content/u...)
I tried this with some NBA draft ranks a few years ago here: gist.github.com/awhug/46d5d3...
The Bayesian bootstrap is probably your best bet for this analysis imo. This package should give you both regular and partial correlations:
github.com/donaldRwilli...
Most of the other Bayesian formulations I’ve seen are a bit awkward in comparison.
github.com/donaldRwilli...
Most of the other Bayesian formulations I’ve seen are a bit awkward in comparison.
GitHub - donaldRwilliams/BBcor: Bayesian Bootstrapped Correlations
Bayesian Bootstrapped Correlations. Contribute to donaldRwilliams/BBcor development by creating an account on GitHub.
github.com
December 26, 2024 at 10:15 AM
The Bayesian bootstrap is probably your best bet for this analysis imo. This package should give you both regular and partial correlations:
github.com/donaldRwilli...
Most of the other Bayesian formulations I’ve seen are a bit awkward in comparison.
github.com/donaldRwilli...
Most of the other Bayesian formulations I’ve seen are a bit awkward in comparison.
I found myself nerd-sniped at the office one day during my psych PhD fitting non-linear lactation models. I had absolutely no business answering this, but they are actually very cool!
stats.stackexchange.com/a/500770/212...
stats.stackexchange.com/a/500770/212...
Fitting model to Wood's lactation curve
I'm trying to determine how I can estimate the following model:
$y = at^bexp(ct)$, where $a$, $b$ and $c$ are estimate from a set of data.
But I can't figure out how to derive the formula into some...
stats.stackexchange.com
December 9, 2024 at 5:10 AM
I found myself nerd-sniped at the office one day during my psych PhD fitting non-linear lactation models. I had absolutely no business answering this, but they are actually very cool!
stats.stackexchange.com/a/500770/212...
stats.stackexchange.com/a/500770/212...
Yeah, no argument from me. My comment was more yelling at the sky from reading too much questionable applied psychometrics.
I guess all I was getting at is that there are in theory very clever ways of uncovering these relationships, but whether they work so well in the real world feels dubious.
I guess all I was getting at is that there are in theory very clever ways of uncovering these relationships, but whether they work so well in the real world feels dubious.
December 4, 2024 at 1:42 PM
Yeah, no argument from me. My comment was more yelling at the sky from reading too much questionable applied psychometrics.
I guess all I was getting at is that there are in theory very clever ways of uncovering these relationships, but whether they work so well in the real world feels dubious.
I guess all I was getting at is that there are in theory very clever ways of uncovering these relationships, but whether they work so well in the real world feels dubious.
This logic is totally sound, but has also been instrumental in the proliferation of analyses ‘corrected for attenuation’ which often seem to make the results opaque and at times frankly implausible.
December 4, 2024 at 6:56 AM
This logic is totally sound, but has also been instrumental in the proliferation of analyses ‘corrected for attenuation’ which often seem to make the results opaque and at times frankly implausible.
After spending this week trying to explain in lay terms what odds ratio outputs mean, I am very prepared to take the risk just go Poisson 😉
(And yes I am aware that marginal effects exist but no they aren’t so useful in my case)
(And yes I am aware that marginal effects exist but no they aren’t so useful in my case)
December 4, 2024 at 5:49 AM
After spending this week trying to explain in lay terms what odds ratio outputs mean, I am very prepared to take the risk just go Poisson 😉
(And yes I am aware that marginal effects exist but no they aren’t so useful in my case)
(And yes I am aware that marginal effects exist but no they aren’t so useful in my case)
Are you familiar with 538’s model?
Because they do geographically model votes and incorporate other external data on a range of economic and political indicators.
abcnews.go.com/538/538s-202...
Do you have specific thoughts on how the model could be improved to be more informative?
Because they do geographically model votes and incorporate other external data on a range of economic and political indicators.
abcnews.go.com/538/538s-202...
Do you have specific thoughts on how the model could be improved to be more informative?
How 538's 2024 presidential election forecast works
The methodology of how 538's 2024 presidential election forecast model works.
abcnews.go.com
October 31, 2024 at 3:02 AM
Are you familiar with 538’s model?
Because they do geographically model votes and incorporate other external data on a range of economic and political indicators.
abcnews.go.com/538/538s-202...
Do you have specific thoughts on how the model could be improved to be more informative?
Because they do geographically model votes and incorporate other external data on a range of economic and political indicators.
abcnews.go.com/538/538s-202...
Do you have specific thoughts on how the model could be improved to be more informative?
What (if anything) is a more informative alternative?
October 31, 2024 at 1:21 AM
What (if anything) is a more informative alternative?
Ok but on the other hand lots of us like you 🥹👉👈
October 30, 2024 at 6:45 AM
Ok but on the other hand lots of us like you 🥹👉👈
Totally and embarrassingly wrong. This means your coefficients are estimated without error. Other scientists could only dream of this precision.
September 27, 2024 at 10:41 PM
Totally and embarrassingly wrong. This means your coefficients are estimated without error. Other scientists could only dream of this precision.