Standard Normal

In fact, it would only take an unmeasured confounder with a HR =0.9 and a 0.2 SD difference in means to tip this result. See

tipr::tip_hr_with_continuous(effect_observed = .98, exposure_confounder_effect = .2)

Sorry, pointing to early separation in prior RCTs does not reassure me that there is no residual confounding in this weighted analysis of observational data. Also sad to see the BMJ hyping the 11% number, failing to point out that (if we believe estimates) the data is consistent with a 2% difference

These are trials? If you'd like me to be more verbose: this kind of early separation in curves in an observational study using PS methods to try and eliminate confounding is most likely due to residual confounding

Curves separate immediately which is a clear sign of residual confounding

Peer review of prediction model studies in oncology needs improvement: a systematic review of open peer review reports from BMC journals - Journal of Clinical Epidemiology www.jclinepi.com/article/S089...

Well unfortunately people are going to do observational studies where RCTs raise ethical issues (e.g. pregnancy). Not having subject matter experts involved increases the chance of bias. Obviously possible for an epidemiologist to gain expertise but probably easier to collaborate

Fascinating. So in a situation where you're trying to determine which of two treatments leads to better outcomes, having no idea of how treatment decisions are made would be just fine as long as you have general epi expertise?

If you want to understand potential confounding factors domain expertise is vital

NYT: Science is broken. Can this man (Jay Bhattacharya) fix it?

No. The answer is no

What @ezraklein.bsky.social would call "politics the right way". Lol

If someone walks away from this understanding that bus usage outside London is down roughly 50% whereas in London it has increased roughly 100% in the same period, where have they been misled? Honestly curious

I genuinely don't understand what is misleading about this?

It would depend on what kind of spline you are using. If you're doing the analysis in R the rms package has functions for printing the model equation

It is baffling to me how many senior researchers make blanket statements degrading other's work. Please educate the group and show us a recent simulation study (e.g. in SiM) where you could do "the calculations" (this might involve linking to a paper that is not your own)

Associations of semaglutide with Alzheimer's disease-related dementias in patients with type 2 diabetes: A real-world target trial emulation study - PubMed BackgroundAlmost half of the dementia cases are preventable. Semaglutide treats several medical conditions that are risk factors for dementia.ObjectiveWe aim to investigate if semaglutide is associate...

If this is a "target trial emulation" then the phrase has no meaning pubmed.ncbi.nlm.nih.gov/40552638/

Casual Inference: Combining Data & Making Effects Generalizable with Carly Brantner | Season 6 Episode 7 Carly Brantner is an assistant professor of Biostatistics & Bioinformatics at Duke University and Duke Clinical Research Institute. Follow along on Bluesky: Carly: Ellie:  Lucy:  &n...

Wrong link casualinfer.libsyn.com/site/combini...

In R
`Hmisc::getHdata("gusto")`

GUSTO-1 is big - see www.fharrell.com/post/path/

Last one

Not sure why one would do both split sample *and* bootstrap for internal validation. Am I missing something?

For those interested in what not to do www.frontiersin.org/journals/car...

Did you just rebrand Master of Health Informatics (MHI) degrees?

External validation of an artificial intelligence model using clinical variables, including ICD-10 codes, for predicting in-hospital mortality among trauma patients: a multicenter retrospective cohort... Scientific Reports - External validation of an artificial intelligence model using clinical variables, including ICD-10 codes, for predicting in-hospital mortality among trauma patients: a...

www.nature.com/articles/s41...

I agree the terminology is contradictory; the 'inferior' is referring to 0 and the 'non-inferior' is referring to the NI margin. If we have to have NI trials your wording would be much better!

figure from https://www.nature.com/articles/s41416-022-01937-w showing interpretation of NI trials

The one word should have really been 'inconclusive'

Not a good look to refer to naturalized citizens as foreigners... Even if he is a menace