Karl Rohe
@karlrohe.bsky.social
NO KINGS
“Overly optimistic” 🦮 in Statistics.
Listening in statistics.
Statistics Professor at UW Madison.
“Overly optimistic” 🦮 in Statistics.
Listening in statistics.
Statistics Professor at UW Madison.
claude: <answers intelligently>
me: what questions would this person ask me to help me decide how to proceed?
me: what questions would this person ask me to help me decide how to proceed?
November 22, 2025 at 3:57 PM
claude: <answers intelligently>
me: what questions would this person ask me to help me decide how to proceed?
me: what questions would this person ask me to help me decide how to proceed?
heck yeah. don't be a stranger if i might help point
its on back burner right now, but i've tried to make an r package to help think-about / express these things more carefully. complete rethinking of how we think about pca.... using a formula like we do for lm(y~...)
karlrohe.github.io/longpca/
its on back burner right now, but i've tried to make an r package to help think-about / express these things more carefully. complete rethinking of how we think about pca.... using a formula like we do for lm(y~...)
karlrohe.github.io/longpca/
A formula interface for model-first PCA; PCA for the people!
This package introduces a novel formula syntax for PCA. In modern applications (where data is often in "long format"), the formula syntax helps to fluidly imagine PCA without thinking about matrices....
karlrohe.github.io
November 21, 2025 at 8:23 PM
heck yeah. don't be a stranger if i might help point
its on back burner right now, but i've tried to make an r package to help think-about / express these things more carefully. complete rethinking of how we think about pca.... using a formula like we do for lm(y~...)
karlrohe.github.io/longpca/
its on back burner right now, but i've tried to make an r package to help think-about / express these things more carefully. complete rethinking of how we think about pca.... using a formula like we do for lm(y~...)
karlrohe.github.io/longpca/
Fixed it for all of us
November 20, 2025 at 6:52 PM
Fixed it for all of us
Are any of the discussants on blue sky???
November 20, 2025 at 6:37 PM
Are any of the discussants on blue sky???
Factor analysis and varimax have been widely misunderstood among statisticians. Two years ago, Muzhe Zeng and I had a discussion paper at JRSS-B that tried to clarify why factor analysis and varimax were actually decades ahead of their time
It’s free (with discussion) here:
doi.org/10.1093/jrss...
It’s free (with discussion) here:
doi.org/10.1093/jrss...
Vintage factor analysis with Varimax performs statistical inference
Abstract. In the 1930s, Psychologists began developing Multiple-Factor Analysis to decompose multivariate data into a small number of interpretable factors
doi.org
November 20, 2025 at 6:28 PM
Factor analysis and varimax have been widely misunderstood among statisticians. Two years ago, Muzhe Zeng and I had a discussion paper at JRSS-B that tried to clarify why factor analysis and varimax were actually decades ahead of their time
It’s free (with discussion) here:
doi.org/10.1093/jrss...
It’s free (with discussion) here:
doi.org/10.1093/jrss...
empirically, there often tend to be "clusters of directions" and these appear as "radial streaks".
This was first noticed by Thurstone, the inventor of Factor Analysis.
heard of “topic modeling”? they generate streaks!
More generally, *varimax* after pca is a great way to find these streaks.
This was first noticed by Thurstone, the inventor of Factor Analysis.
heard of “topic modeling”? they generate streaks!
More generally, *varimax* after pca is a great way to find these streaks.
November 20, 2025 at 6:28 PM
empirically, there often tend to be "clusters of directions" and these appear as "radial streaks".
This was first noticed by Thurstone, the inventor of Factor Analysis.
heard of “topic modeling”? they generate streaks!
More generally, *varimax* after pca is a great way to find these streaks.
This was first noticed by Thurstone, the inventor of Factor Analysis.
heard of “topic modeling”? they generate streaks!
More generally, *varimax* after pca is a great way to find these streaks.
then it becomes natural to think in "polar coordinates" (not Euclidean).
in polar coordinates, there is a direction and a radius. When the origin is special, is helpful to separate these two things.
There is some wild “history of data analysis” here…
in polar coordinates, there is a direction and a radius. When the origin is special, is helpful to separate these two things.
There is some wild “history of data analysis” here…
November 20, 2025 at 6:28 PM
then it becomes natural to think in "polar coordinates" (not Euclidean).
in polar coordinates, there is a direction and a radius. When the origin is special, is helpful to separate these two things.
There is some wild “history of data analysis” here…
in polar coordinates, there is a direction and a radius. When the origin is special, is helpful to separate these two things.
There is some wild “history of data analysis” here…
How about “October rain” ?
November 18, 2025 at 8:06 PM
How about “October rain” ?
like... jeeze claude.
November 18, 2025 at 6:29 PM
like... jeeze claude.
request that was rejected for safety reasons:
when user clicks view template, we get a dialogue box... the colors are bad for the template.... not enough contrast, etc. i want to match the same color format as is used on <different page url> ... can you see that coloring?
when user clicks view template, we get a dialogue box... the colors are bad for the template.... not enough contrast, etc. i want to match the same color format as is used on <different page url> ... can you see that coloring?
November 18, 2025 at 6:29 PM
request that was rejected for safety reasons:
when user clicks view template, we get a dialogue box... the colors are bad for the template.... not enough contrast, etc. i want to match the same color format as is used on <different page url> ... can you see that coloring?
when user clicks view template, we get a dialogue box... the colors are bad for the template.... not enough contrast, etc. i want to match the same color format as is used on <different page url> ... can you see that coloring?
also, if anyone wants to take be deer hunting in southern wisconsin, that would be fun.
November 10, 2025 at 11:07 PM
also, if anyone wants to take be deer hunting in southern wisconsin, that would be fun.
I’m so sorry. Sending you both comfort and well wishes.
November 6, 2025 at 12:45 AM
I’m so sorry. Sending you both comfort and well wishes.
Summary from claude:
November 2, 2025 at 12:27 PM
Summary from claude:
Is this like “clothing optional”?
October 31, 2025 at 7:29 PM
Is this like “clothing optional”?