Norm Matloff (你有冇諗清楚呀?)

Agree 100%.

I never use them myself, but have no opinion, strong or weak, about them.

Here is the author list:

Paul Murrell;
Hadley Wickham;
Ross Ihaka;
Kurt Hornik, Uwe Ligges;
Norm Matloff;
Emi Tanaka;
Baddeley, Ege Rubak, Rolf Turner
Weihao Li, Dianne Cook, Emi Tanaka,
Susan Vanderplas, Klaus Ackermann

3/3

The special issue is not quite ready yet, but here is the temporary link to my article:

onlinelibrary.wiley.com/share/author...

I welcome your comments. 2/

#rstats

The Australian and New Zealand Journal of Statistics will be running a special issue commemorating the 25th anniversary of the official release of R. (The Univ. of Auckland was the birthplace of R.) They invited various people to contribute articles, including me. 🧵 1/

Interesting thread, not just about tibbles but general issues of inheritance.

Very nice, esp. multiple models side-by-side.

Great writeup, should be required reading for students in Data Science and Statistics.

Super presentation! But I'd also urge you to look at

matloff.github.io/No-P-Values/...

Though I do strongly encourage those who know only R to learn Python (see my old tutorial, heather.cs.ucdavis.edu/matloff/publ...), IMO this should not be done until the person is VERY skilled in R.

Plus...I just don't see ANY benefit for me from IDEs. I've tried quite a few BTW.

Please note: I do not use ANY IDE. I don't use IDEs for C/C++, Python, LaTeX, Quarto etc.

IDEs take up precious space on my screen. They cannot be used for debugging parallel code. Typing complex math expressions is SLOWER than writing them with my Vim macros.

RStudio is good for coder novices.

Querying India's MoSPI Data with Claude and MCP | Aman Bhargava The Ministry of Statistics and Programme Implementation released a new tool for LLM-assisted querying of their survey data. I explore some of it's possibilities.

The Indian government released an MCP server that lets you query their survey data using plain English and get back aggregate statistics from national surveys. I explore some of its possibilities and limitations. #data #stats

aman.bh/blog/2026/qu...

For us Vim fans, the real advantage is programmability, not key bindings. But neither Rstudio nor any other IDE I'm aware of allows this.

startup idea: wearable device app for automatic medication adherence with machine learning based reminders

Good point, true for a lot of things.

Interesting idea, nice graphics. Does one have control over the number of bootstrap replications?

Very nice! Any installation/configuration problems?

The decision is "arguably correct" if the prior is "arguably correct."

Depends on what we mean by "fit." One definition, which you are implicitly using would integrated (squared, absolute etc.), but there are others, such as those focusing on sharpness of the peak.

Not sure what you meant by references to AOC and the GOP, but you may find tinyurl.com/5fpkhune of interest.

Unfortunately, this "two moving targets" setting has been the standard genre for math stat people over the last couple of decades. It is a natural approach given the situation, and makes for impressive mathematics, but it is not of much practical value, IMO.

Actually, a number of papers have been published that purport to enable inference, but yes, they are "tricks." It all depends on how one chooses lambda, and how fast lambda -> 0 in terms of n, which IMO is meaningless in practice. 🧵 1/

What’s one data tool or concept you rediscovered in 2025 that surprised you with its usefulness? For me: data.table in R. Lightning-fast joins. Elegant logic.

#RStats #DataScience #MachineLearning

This is the public's web page, not the GOP's.

Regularization, no. One cannot do statistical inference there.

Trimmed memes, yes, provided one recognizes that the quantity to be estimated has now changed. But even there, the goal is to get rid of bad data, not to incorporate it!

Yes, the estimation vs. prediction issue is very important. In terms of the "possible impact on the public" criteria I mentioned (in the concurrent thread on X),estimation is the main concern.

So we see that the Bayesian approach devalues our data.

I once even saw a well-credentialed analyst here in Bluesky say, "An advantage of Bayes is that it gives you the probability..." From context, esp. the use of the word "the," I doubt that they understood either. I've seen others like this. 🧵 4/4

Note that that doc says an advantage of Bayes is that they "provide probabilistic insights that enhance decision-making under uncertainty." No mention is made of the fact that those "probabilities" are contrived, not at all like what unsuspecting non-techie readers think of as probabilities. 🧵 3/