dsmccormick.bsky.social
Neophyte #rstats practitioner; infographic #datavis aesthete; cyclist; dog lover
53 posts
61 followers
90 following
Active Commenter
comment in response to
post
@juliasilge.com that was a brilliant tutorial. Loved it. Thanks for the clever use of the dataset to learn poisson regression.
comment in response to
post
Is what you are trying to express a time lag between, arrival, admission, and subsequent specialist interaction? Are there distributions of time lags by time of day? day of week? Is there a time lag distribution that you could visualize? Could you group by day of week and show the distributions?
comment in response to
post
…so I think that would have been helpful to point out, because I was scratching my head with the dataframe example for a couple of minutes, until I read the sample() documentation and saw you can provide *an integer*, vector, or other type to the function.
comment in response to
post
Thanks for the random sampling post. I think there is a subtlety to sample() that you didn’t explain, namely that if you invoke sample(n, num_samples), will get random samples of an integer vector from 1:n. That’s how your dataframe sampling trick works with nrow(df), which is an integer.
comment in response to
post
And I thought of the great Jon Bon Neumann quote: “Anyone who attempts to generate random numbers by deterministic means is, of course, living in a state of sin.”
comment in response to
post
That was a complete tour de force of yak shaving.
comment in response to
post
Thanks for pointing that out. I think the axis rescaling by facet_wrap() made the log scaled axis too cluttered. I’ll have to find a workaround.
comment in response to
post
…but I was really trying to get rid of the *text* labels for minor breaks.
comment in response to
post
I was using a scales-based anonymous function to create the logarithmic minor break lines. I’ll try your suggestion. Thanks.
comment in response to
post
This was so timely! I was literally trying to redo a function with multiple values to return. Thanks for the comprehensive treatment.
comment in response to
post
I agree that when I started using JetBrains Mono for my coding font, those glyphs make it very readable. So good point.
comment in response to
post
Why? It seems to work in every other programming language and its used in function argument assignments just fine.
comment in response to
post
Jerusalem.
comment in response to
post
This reminds me of Richard Hofstadter’s en.wikipedia.org/wiki/The_Par...
comment in response to
post
I don’t currently have processing bottlenecks that merit considering polars or duckdb, but it’s astonishing that you get literally 1 and 2 orders of magnitude performance improvement with those packages. Thank you for the example.
comment in response to
post
Ah, the client….
comment in response to
post
There are arrow and polars R packages. But polars is probably much more evolved in Python since that was the original development target.
comment in response to
post
The thing that was particularly cool was how you give the LLM the data context as well as the source code context for more accurate code completion. So clever.
comment in response to
post
That was an excellent interview. Very eye opening as to what you have accomplished with pal and gander and looking forward to having those tools available in the IDE.
comment in response to
post
Why is statistical significance the important issue? What is the question you want to answer? Why not compare evidence for competing models with a Bayesian analysis, even if the other model is some version of a null hypothesis?
comment in response to
post
would this be of any use? gradientdescending.com/select-colou...
comment in response to
post
And to be a little less glib, you now really understand the problem you were trying to solve.
comment in response to
post
You are getting a discount for your future self who won’t make that same mistake again.
comment in response to
post
Actually, the function I needed had to grab the column names as well, which does require the more involved map_chr() function calls. But I’m so glad to learn about unite().
comment in response to
post
OMG I just tried out unite() and it does precisely what I wanted in one function call. I didn’t know about unite(). Thank you !
comment in response to
post
I have a purrr version of your column string concatenation if you are interested. It allows for picking arbitrary lists of columns. I had to go to generative AI for the answer because I’m not good at functional programming yet.
comment in response to
post
Terrific resource that you have authored. Thanks very much.
comment in response to
post
I saw that as well, so some hope for ongoing public access.
comment in response to
post
You can also get ChatGPT or related services to make you a shiny app skeleton. Just specify what’s in your data frame and what shiny UI elements you want. It’s pretty quick to start. I really like working with leaflet maps.
comment in response to
post
But things like public access to Census Department data for things like geographic data. This is public data that the public paid for collection and distribution. It’s absolutely outrageous.
comment in response to
post
I just tried and it’s still blocked. This is so messed up. It’s the public’s data.
comment in response to
post
This reminds me of the tv show Person Of Interest. It all seems so dystopian to me.
comment in response to
post
And as you describe the experience, gave me the perspective to weather just about anything.
comment in response to
post
There were times of poignant beauty raised in high relief because of mortality. And there were many times of glorious laughter and happiness with friends, our children, our family. It was brutal at times, but it was incredibly beautiful and meaningful.
comment in response to
post
I was on the opposite side: my wife’s cancer killed her 3.5 years ago. It was 25 years with 5 serious ones at the end. I concur that our tempo of attention was as you say: do the medical things, then live.
comment in response to
post
I wasn’t sure. Thanks for clarifying.
comment in response to
post
Have you seen JASP: jasp-stats.org it has both Bayesian and frequentist tests. It’s free.
comment in response to
post
Have you heard about JASP? It’s free and has both Bayesian and frequentist tests: jasp-stats.org
comment in response to
post
Love it. I had one of those in grad school.
comment in response to
post
Is that an original Macintosh? I owned several generations of those wonderful Macs.
comment in response to
post
3/ The book itself is beautiful and goes nicely with any of the Tufte books. Thanks, Kieran.
comment in response to
post
2/ One thing I liked is how @kjhealy.co included examples of the subtleties of ggplot, like how the fill color doesn’t go in the mapping (ie, aes()) in layers like geom_col(); he showed how things could go wrong as well as how to fix them.
comment in response to
post
I’m relatively new (2 years) to R, but the combination of the pipe operators and vector operations make for such compact and comprehensible actions. It almost seems like cheating at times what I can do I a few lines of R code.
comment in response to
post
I never made that connection to Magritte. Thanks for that. Tres recherché.
comment in response to
post
I am reading @rmkubinec.bsky.social ‘s Bayesian Hitman this week. And that was a good episode.
comment in response to
post
I have gotten a lot out of Learning Bayesian Statisics (learnbayesstats.com) and Super Data Science (www.superdatascience.com/podcast).
comment in response to
post
Who are you quoting?