Norm Matloff (你有冇諗清楚呀?)
@matloff.bsky.social
Em. Prof., UC Davis. Many awards, incl. book, teaching, public service. Many books, latest The Art of Machine Learning (uses qeML pkg). Former Editor in Chief, the R Journal. Views mine. heather.cs.ucdavis.edu/matloff.html
Seems to me -- for now, at least -- Grok would not force the notion of IQ (a notion I don't believe in) into most technical discussions. I use Grok in addition to all the other major ones, and haven't seen it do anything like this.
October 26, 2025 at 6:43 PM
Seems to me -- for now, at least -- Grok would not force the notion of IQ (a notion I don't believe in) into most technical discussions. I use Grok in addition to all the other major ones, and haven't seen it do anything like this.
No algorithm will be perfect, but for technical issues I doubt that the new policy will make much difference to me. Meanwhile, the volume on X is high, while it is too low on technical issues here on BlueSky. I follow both platforms.
October 26, 2025 at 6:31 PM
No algorithm will be perfect, but for technical issues I doubt that the new policy will make much difference to me. Meanwhile, the volume on X is high, while it is too low on technical issues here on BlueSky. I follow both platforms.
On the off chance that someone reads this old thread: My references to "Larry W" should be "Ron W," i.e. Ron Wasserstein, Executive Director of the ASA, not stat professor Larry Wasserman. Sorry for the confusion.
October 24, 2025 at 2:50 PM
On the off chance that someone reads this old thread: My references to "Larry W" should be "Ron W," i.e. Ron Wasserstein, Executive Director of the ASA, not stat professor Larry Wasserman. Sorry for the confusion.
Yes, there are many possible measures. I'm not defending the variance-based one, just explaining for the OP the motivation for its general use in classical statistics.
October 24, 2025 at 1:58 PM
Yes, there are many possible measures. I'm not defending the variance-based one, just explaining for the OP the motivation for its general use in classical statistics.
Variance is a natural measure, since it determines (for large samples) the widths of confidence intervals and powers of tests.
October 24, 2025 at 5:45 AM
Variance is a natural measure, since it determines (for large samples) the widths of confidence intervals and powers of tests.
Excellent work by two top researchers!
October 22, 2025 at 10:23 PM
Excellent work by two top researchers!
Unfortunately, students want formulaic answers, especially in programs like Data Science that promise them jobs. Questions like "Do you have the right data?" are very un-formulaic, exactly what the students DON'T want, so if you can get them to meaningfully address such issues, my hat is off to you.
October 17, 2025 at 4:30 PM
Unfortunately, students want formulaic answers, especially in programs like Data Science that promise them jobs. Questions like "Do you have the right data?" are very un-formulaic, exactly what the students DON'T want, so if you can get them to meaningfully address such issues, my hat is off to you.
Ah, yes, thanks!
October 17, 2025 at 4:40 AM
Ah, yes, thanks!
The data.table package should be in there somewhere.
October 15, 2025 at 9:08 PM
The data.table package should be in there somewhere.
For one thing, the area in question is often infinite. Second, it's more of a conceptual thing than something that is often used in daily work. Third, why make it harder?
October 14, 2025 at 4:25 AM
For one thing, the area in question is often infinite. Second, it's more of a conceptual thing than something that is often used in daily work. Third, why make it harder?
Great title, Joe! I might suggest a slight change: Keeping Extremist Advocates of LLMs in Their Lane. :-)
October 12, 2025 at 5:55 PM
Great title, Joe! I might suggest a slight change: Keeping Extremist Advocates of LLMs in Their Lane. :-)
Reposted by Norm Matloff (你有冇諗清楚呀?)
A friend of mine was hired at a company which uses their own self-hosted AI tools a few weeks ago.
Just today they've told me: "Dude, I think I'm only gonna last 2 months here. Their energy bills make it financially insolvent."
Just today they've told me: "Dude, I think I'm only gonna last 2 months here. Their energy bills make it financially insolvent."
October 8, 2025 at 4:06 PM
A friend of mine was hired at a company which uses their own self-hosted AI tools a few weeks ago.
Just today they've told me: "Dude, I think I'm only gonna last 2 months here. Their energy bills make it financially insolvent."
Just today they've told me: "Dude, I think I'm only gonna last 2 months here. Their energy bills make it financially insolvent."
Elegant code is often more difficult to write, with subtler bugs, and difficult to debug. In the case of R, it might also run faster, but there is a tradeoff here, no consistent winner in all cases.
October 8, 2025 at 11:17 PM
Elegant code is often more difficult to write, with subtler bugs, and difficult to debug. In the case of R, it might also run faster, but there is a tradeoff here, no consistent winner in all cases.
After years of hearing it, I still don't understand why stepwise is considered so evil by some.
October 7, 2025 at 5:26 AM
After years of hearing it, I still don't understand why stepwise is considered so evil by some.
IMO, it's absurd to try to interpret a 4-way interaction. Even explaining 3-ways is questionable.
October 6, 2025 at 8:08 PM
IMO, it's absurd to try to interpret a 4-way interaction. Even explaining 3-ways is questionable.
Thanks, looks like great stuff on GMM, a topic I've always wanted to pursue but (like many others) have never had the time to do much.
October 3, 2025 at 9:02 PM
Thanks, looks like great stuff on GMM, a topic I've always wanted to pursue but (like many others) have never had the time to do much.
Thanks, nice work, looking forward to the preprint.
October 3, 2025 at 8:01 PM
Thanks, nice work, looking forward to the preprint.
I strongly agree. The function gtools::defmacro is very important in my work. (I'll give examples here is anyone is interested.) If you have some short misc. task to do, check gtools before writing your own.
October 2, 2025 at 11:35 PM
I strongly agree. The function gtools::defmacro is very important in my work. (I'll give examples here is anyone is interested.) If you have some short misc. task to do, check gtools before writing your own.
Nice! But what is the theory behind it?
October 1, 2025 at 3:32 AM
Nice! But what is the theory behind it?
I certainly never noticed it. Good catch!
September 27, 2025 at 7:44 PM
I certainly never noticed it. Good catch!
What is the value of epsilon here?
September 27, 2025 at 5:37 PM
What is the value of epsilon here?