Solomon Kurz
@solomonkurz.bsky.social
Clinical psychology researcher | applied statistics geek | so called #RStats influencer
This sums up my post-Phd career, so far:
September 29, 2025 at 9:10 PM
This sums up my post-Phd career, so far:
Here you go. Totally normal result. Happens all the time.
August 12, 2025 at 2:31 PM
Here you go. Totally normal result. Happens all the time.
The difference between the two isn't trivial. Using Kruschke initial IQ data example, here's the two versions of d, based on standardizer:
August 6, 2025 at 4:15 PM
The difference between the two isn't trivial. Using Kruschke initial IQ data example, here's the two versions of d, based on standardizer:
Cohen's d was explicitly developed with the Gaussian framework. To avoid confusion, what would someone call an SMD from a Student-t model, and how would one differentiate between one standardized with the pooled SD versus the pooled scale? Do these look okay? What other recommendations do you have?
August 6, 2025 at 3:21 PM
Cohen's d was explicitly developed with the Gaussian framework. To avoid confusion, what would someone call an SMD from a Student-t model, and how would one differentiate between one standardized with the pooled SD versus the pooled scale? Do these look okay? What other recommendations do you have?
Found it: theoatmeal.com/comics/desig...
August 5, 2025 at 5:35 PM
Found it: theoatmeal.com/comics/desig...
For reference, here's what such an effect size would look like on the z-scale metric:
July 26, 2025 at 12:39 AM
For reference, here's what such an effect size would look like on the z-scale metric:
But these data have some outliers (by design), and if we fit a robust Student-t t model to them instead (which fits the data very well), the posterior means get jerked around, though the medians are still pretty close:
June 18, 2025 at 2:39 PM
But these data have some outliers (by design), and if we fit a robust Student-t t model to them instead (which fits the data very well), the posterior means get jerked around, though the medians are still pretty close:
If you just focus on central tendencies, they sometimes approximate one another very closely. In this model, this holds for posterior means and medians.
June 18, 2025 at 2:39 PM
If you just focus on central tendencies, they sometimes approximate one another very closely. In this model, this holds for posterior means and medians.
First, to make my point clear (though I suspect you get it), residuals and predictive errors have full posterior distributions. Here's an example from a Gaussian model:
June 18, 2025 at 2:39 PM
First, to make my point clear (though I suspect you get it), residuals and predictive errors have full posterior distributions. Here's an example from a Gaussian model:
* solving the HMC proposal schematics in Chapter 14, and
* moving closer to official tidyverse style formatting.
This is a transitional update. More are likely to come over the next year, with some pretty cool additions. But we had to start somewhere, so enjoy incremental progress for now.
3/3
* moving closer to official tidyverse style formatting.
This is a transitional update. More are likely to come over the next year, with some pretty cool additions. But we had to start somewhere, so enjoy incremental progress for now.
3/3
April 14, 2025 at 3:01 PM
* solving the HMC proposal schematics in Chapter 14, and
* moving closer to official tidyverse style formatting.
This is a transitional update. More are likely to come over the next year, with some pretty cool additions. But we had to start somewhere, so enjoy incremental progress for now.
3/3
* moving closer to official tidyverse style formatting.
This is a transitional update. More are likely to come over the next year, with some pretty cool additions. But we had to start somewhere, so enjoy incremental progress for now.
3/3
Some noteworthy changes include:
* switching to quarto book,
* adopting the base pipe |>,
* fixing some mistakes in Chapters 4 and 10,
* updating the workflows in Chapters 5 and 6 to be more faithful to the spirit of the text,
2/3
* switching to quarto book,
* adopting the base pipe |>,
* fixing some mistakes in Chapters 4 and 10,
* updating the workflows in Chapters 5 and 6 to be more faithful to the spirit of the text,
2/3
April 14, 2025 at 3:01 PM
Some noteworthy changes include:
* switching to quarto book,
* adopting the base pipe |>,
* fixing some mistakes in Chapters 4 and 10,
* updating the workflows in Chapters 5 and 6 to be more faithful to the spirit of the text,
2/3
* switching to quarto book,
* adopting the base pipe |>,
* fixing some mistakes in Chapters 4 and 10,
* updating the workflows in Chapters 5 and 6 to be more faithful to the spirit of the text,
2/3
3D scatterplots are not natively supported {ggplot2}. Who has good workarounds?
I'm looking to make plots like this with a method that doesn't make me want to throw myself into the sea:
#rstats
I'm looking to make plots like this with a method that doesn't make me want to throw myself into the sea:
#rstats
March 17, 2025 at 9:27 PM
3D scatterplots are not natively supported {ggplot2}. Who has good workarounds?
I'm looking to make plots like this with a method that doesn't make me want to throw myself into the sea:
#rstats
I'm looking to make plots like this with a method that doesn't make me want to throw myself into the sea:
#rstats
I believe this would be the DAG:
3/3
3/3
February 18, 2025 at 6:39 PM
I believe this would be the DAG:
3/3
3/3
Figure 1 from doi.org/10.3758/s134...
February 13, 2025 at 4:31 AM
Figure 1 from doi.org/10.3758/s134...
I made a similar plot using Efron and Morris’s classic paper, "Stein’s paradox in statistics," as an example. Here's the full blog post: solomonkurz.netlify.app/blog/2019-02...
January 25, 2024 at 3:14 PM
I made a similar plot using Efron and Morris’s classic paper, "Stein’s paradox in statistics," as an example. Here's the full blog post: solomonkurz.netlify.app/blog/2019-02...
Assuming the percents follow a beta distribution, here's what those sample statistics imply (very orderly data):
November 21, 2023 at 8:33 PM
Assuming the percents follow a beta distribution, here's what those sample statistics imply (very orderly data):
Feeling some love for how flextable::compose() lets me include LaTeX symbols in my #rstats tables:
September 27, 2023 at 8:20 PM
Feeling some love for how flextable::compose() lets me include LaTeX symbols in my #rstats tables: