Washington Irving

Regression modeling strategies by @f2harrell.bsky.social is amongst the best out there, imo.

One of my all time favorite comedic dyads.

Just amazing on so many levels.

As an aside, I thought I was gonna get fired once when I casually added this paper to a meeting chat where some big time clinical researches were presenting some data and the working paper title was ‘finally, an externally validated sepsis prediction algorithm’ or something along those lines 🤦‍♂️

There is no such thing as a validated prediction model - BMC Medicine Background Clinical prediction models should be validated before implementation in clinical practice. But is favorable performance at internal validation or one external validation sufficient to claim that a prediction model works well in the intended clinical context? Main body We argue to the contrary because (1) patient populations vary, (2) measurement procedures vary, and (3) populations and measurements change over time. Hence, we have to expect heterogeneity in model performance between locations and settings, and across time. It follows that prediction models are never truly validated. This does not imply that validation is not important. Rather, the current focus on developing new models should shift to a focus on more extensive, well-conducted, and well-reported validation studies of promising models. Conclusion Principled validation strategies are needed to understand and quantify heterogeneity, monitor performance over time, and update prediction models when appropriate. Such strategies will help to ensure that prediction models stay up-to-date and safe to support clinical decision-making.

Yeah, main issue seems to be that everyone can get their hands on some data and can publish a model. I mean, that was part of what I did for my PhD. I'm not entirely sure the problem is just validation though, I tend to follow van Calster et al's thinking on this: doi.org/10.1186/s129...

100%. And the issues raised in this paper (which is great) get to the core if the problem. But very little is done to combat such problems. So we get more and more decision aids, algorithms, etc. but more or less static sepsis mortality in the last 2 decades

I dunno, having worked in this field for 10+ yrs, I see more and more algorithms published, but less and less validation. This is a problem in medical research, imo—ppl want to derive and publish their own approach as opposed to validating someone else’s (when that’s even possible).

Right. And given the volume of information flowing at clinicians, it doesn’t take much poor performance from the alerts for them to become a nuisance and thus ignored.

Oh for sure, it’s way more than alert fatigue. There are so many interests in heath care, often working at cross purposes.

Plus, sepsis is a heterogeneous disease and a great deal of existing research essentially ignores this, as if sepsis is clear entity that you either have or don’t have.

Also, test characteristics for automated sepsis alerts are awful.

It’s widely recognized amongst clinicians that alert fatigue is real.

There are medical journals, even high impact ones, filled to the brim with papers drawing conclusions based questionable or even demonstrably incorrect statistical methods. And no one seems to mind.

This keeps me up at night.

It’s wrong on so many levels.

But also kind of amazing.

But mainly awful.

I’ll see your lazy and raise you a frightening:

Stepwise variable selection in frequentist framework. Drop those vars into a Bayesian model, declare model ‘correct’ because varying priors doesn’t alter the posterior much.

YouTube video by Rt Hon Francis Urquhart The Simpsons: Ray Patterson reinstated

I fantasize that one day, after being labeled ‘uncooperative’, that I get to return and make this speech.

m.youtube.com/watch?v=RJax...

me, today: there was a mistake in the primary key and they resent the data files.

my brain from 1wk ago: sucker! bwaaahahahahhahaha

my brain, 1wk ago: i know you're importing, modifying, and merging 300MB of pipe-delimited text files. and i know you don't work with multi datetime formats within the same files that often. but you def DONT need to comment your code about what you did and why. you def only do this task ONCE

YouTube video by mikeford Looks like those clowns in Congress did it again...

www.youtube.com/watch?v=IC3W...

'What is your research question?

- a sardonic new drama about a group of jaded medical statisticians who each week must explain the same issues over and over again to their arrogant clinical colleagues. All while receiving minimum recognition or reward.

I'd watch that!

Hank Hill I do not want to save to onedrive. I want to save to the documents folder... On my computer... That I own... In my house.

a bit...maybe. it seems that it *might* be possible to extract this info fiddling with the stan code. alternatively, recreating the plots manually by sampling from the posterior draws...which seems more in line with current skills. although not totally sure this will work

I’ve never totally understood why percent decrease is so prevalent. Maybe cause it’s easier to show a ‘large’ difference on a relative scale?

He’s def not Harrison-fording in it.

A noble spirit embiggens the smallest man

The combo of Harrison ford and tommy lee jones is top top. I bet ive seen this movie 50 times.

Where does ‘regarding Henry’ fit in the pantheon?

Ok, yeah. I see what you mean. Different sort of roles.

You mean besides Star Wars trilogy?

One of my favorite anecdotes from THE PREHISTORY OF THE FAR SIDE: "That doesn't sound like the Jane Goodall we know."

YouTube video by apala734 WHY?! Why was I programmed to feel pain?

www.youtube.com/watch?v=nQ-g...

I love when I see people whom I believe to be smarter than I am vocalizing their own uncertainty and need to be in a constant state of learning