Frank Harrell

Biostatistician/Professor/Consultant/Founding Chair of Biostat, Vanderbilt U. Blog: Statistical Thinking: @f2harrell on

[Frank Harrell @f2harrell]

I'd like to give this a proper response rather than a quick one, and I'll make time over the coming week. One thing first, so that I'm engaging your position and not an approximation of it: since this argument came from Gemini in response to your prompt, can you confirm you endorse it as your own? If you'd rather I take up something you've written directly, point me to it and I'll do that instead. If I don't hear back I'll take the post at face value and respond to it as your argument.

[Frank Harrell @f2harrell]

@yudapearl Far better would be to have the first course be on decision making, study design, and measurement properties. Very little controversy and lifelong lessons that are easy to remember.

[Frank Harrell @f2harrell]

@5_utr @TonyFelefly @DavidKrugMD This is a great example and it's from @stephensenn

[Frank Harrell @f2harrell]

@ildiazm Personal attack = claim that I'm inept and need you to do my work. Read what you wrote.

[Frank Harrell @f2harrell]

@ildiazm I thought you had gotten past personal attacks. This attitude does not serve you well Iván.

[Frank Harrell @f2harrell]

@ildiazm Thank you about 'good points'. Please make comments there so you can expand.

[Frank Harrell @f2harrell]

@Think1ing That's the goal :-) , plus to have the Bayesian version of rms::orm (rmsb::blrm) have all the same features. Just not multi-level random effects or random slopes (the ordinal package can do those). #rstats

[Frank Harrell @f2harrell]

Believe it or not these are all special cases of the cumulative probability family of ordinal regression models as discussed in https://t.co/j951pZ9e8z #Statistics

[Frank Harrell @f2harrell]

@ildiazm @JannisSagonasMD Also note that although datamethods users can use nicknames, clicking on them reveal real identities.

[Frank Harrell @f2harrell]

@ildiazm @JannisSagonasMD I don't agree with everything posted on the site. This doesn't fall into the marketing, personal attack, or nothing related to data or public health categories that would cause a moderator to remove the post. Retorts to it are welcomed. BTW hope2respond to your challenge today

[Leena-Palestine @Leena53022916]

*BRILLIANT*
The French made a short film about what could happen after all immigrants would follow the ‘_white, angry old man’s_’ advice to go back where they came from.*

[Frank Harrell @f2harrell]

In the upcoming release of the rms package, the orm function now handles two kinds of random effects structures. So one R function can do every test and fit every model in the table.

[Frank Harrell @f2harrell]

One R function, orm (in the rms package), will fit every model except when using random effects for paired data. The blrm function in the rmsb package handles that.

[Frank Harrell @f2harrell]

@eyayaw @zeddotdev @CotEditor @typstapp Cot's default behavior of flashing a matching [({})], click a delimiter to have Cot highlight all the text until its match, and the wonderful dialog box for find.

[Frank Harrell @f2harrell]

I've been going back and forth between @zeddotdev and @CotEditor and have found that I prefer Cot when not using AI. For the one language I needed syntax highlighting for (@typstapp ) that was not built-in to Cot, Claude quickly wrote the Cot plug-in needed.

[Frank Harrell @f2harrell]

I have a new favorite editor/IDE: Zed. Open-source, blazing fast, built-in AI operability, easy to use. Heard that lots of developers are dropping Visual Studio Code for Zed. Has R and Quarto extensions. https://t.co/SjjChv6jcY

[Frank Harrell @f2harrell]

@pash22 @PulmCrit @NIHRresearch @JelenaSavovic @AndrewFoy82 @drjohnm @TheSGEM @GSCollins @MaartenvSmeden "Clinical risk scores" in how they are usually defined should be replaced with validated risk or life expectancy estimates, and many "risk scores" have been dumbed down by dichotomization. Many of them can be beat re:predictive discrimination using continuous age alone.

[Frank Harrell @f2harrell]

@FamedCelebrity There is only one counter-argument (against p-values not being useful) that works: I was in a hurry and didn't have time to do a Bayesian analysis. I'm still looking for an application where data extremeness assuming theta=0 is the best way to quantify evidence that theta>0.

[Frank Harrell @f2harrell]

@stuartbuck1 I think we've miscast the multiplicity problem all along in an attempt to avoid non-null and compound assertions: https://t.co/pEtwUVCnvA

[Frank Harrell @f2harrell]

@FamedCelebrity I didn't have time to watch much of it but I echo your conclusion. During my multi-decade practice of #Statistics I've never seen a good, compelling use of a p-value. I view p-values as conveniently calculable quantities before we had computers. They were temporarily helpful.

[Frank Harrell @f2harrell]

@ildiazm There are times when that is a reasonable request and times when it's not, related to our previous "statistics is more than math" discussions. And I hope to learn something new too. At a minimum let's hope that readers will see a balanced discussion of both sides.

[Frank Harrell @f2harrell]

@ildiazm You are setting yourself up so that you can't learn from me. I'm trying to set myself up so that I can learn from you.

[Frank Harrell @f2harrell]

@ildiazm You are converting your bias into a future restriction on dialog. Not good. Many things to not require formalism or are benefitted from it.