Devin Incerti retweeted
Uncovering immortal time bias when fitting prognostic survival models in a large clinico-genomic database. Here, our assessment w/ @DevinIncerti, @robtibshirani & all (pre-print).
medrxiv.org/content/10.1101/…
Devin Incerti retweeted
These private equity-owned hospital staffing companies are cutting staff hours as non-#Covid19 business slows:
* Envision ($KKR)
* TeamHealth (Blackstone $BX)
* SCP Health ($ONEX)
* US Acute Care Solutions (Welsh Carson)
* Alteon Health (New Mountain & Frazier)
via @iarnsdorf
Why are ER doctors being sidelined when some hospitals are overflowing? The pandemic is breaking the economics of American health care. propublica.org/article/overw…
Devin Incerti retweeted
NEJM, what are you doing?
Now is not the time to engage in dichotomania of statistical significance.
The conclusions (no significant difference) is not compatible with the data (HR for improvement 1.24; 0.90 to 1.72)
Cc @f2harrell @ADAlthousePhD
nejm.org/doi/full/10.1056/NE…
Devin Incerti retweeted
I've been working on a total rewrite of dtplyr, which lets you take advantage of (almost all of) data.table's speed while using dplyr syntax. The package is still in development, but I'd love for you to try it out and let me know how it goes: dtplyr.tidyverse.org #rstats
Devin Incerti retweeted
Good morning #healtheconomics researchers. We’re accepting proposals to estimate patient preference weights for MCDA using our open-source rheumatoid arthritis model. $50k in funding available. Deadline is 6/15!
thevalueinitiative.org/ivi-a… …
Devin Incerti retweeted
When optimizing decisions you need to maximize per-patient probabilities of making correct choices. That's why individual level ICERs are all-important. Optimizing individual decisions optimizes group outcomes.
Devin Incerti retweeted
This paper motivates the call for the end of significance. A 25% mortality reduction, but because P=0.06 (two-sided), they declare it ‘did not reduce’ mortality. Appalling. jamanetwork.com/journals/jam…
Devin Incerti retweeted
Advertising our short-course “Reproducible Tools for Real-World Data Analyses” at ISPOR May 18. We’ll cover best practices for functional programming, version control, style guides, and peer code review systems. Considering the switch from SAS or Stata to R or Python? Be there.
Devin Incerti retweeted
Stop everything! 🛑 You mean I no longer have to have multiple RStudio instances open at the same time when I get impatient?! 😱
The new Jobs pane in the @rstudio 1.2 Preview lets you run your R scripts in the background. Learn how to use it: blog.rstudio.com/2019/03/14/… #rstats
Devin Incerti retweeted
New @bizScienc Learning Lab - Intro to #MachineLearning in #R
I’ll be demo’ing the new #parsnip package - like #scikitlearn but for #rstats - makes ML in R really easy!
Register here: https://teapawt.org/t.co/iWkcZDwFb2
Devin Incerti retweeted
JAMA rejected this letter from my colleagues & me ("low priority"), so we're publishing on twitter, hoping JAMA will take it more seriously.
Devin Incerti retweeted
We believe in the power of the comment period. @IVI_Health encourages public participation as it can bring attention to valuable evidence or insights that might get overlooked. We invite your comment on our open-source non-small lung cancer value model.
https://teapawt.org/t.co/UotXff2DGj
Devin Incerti retweeted
Clinical trial statistical thought of the day: clinical researchers who truly believe in treatment effect heterogeneity should show their commitment by multiplying sample sizes by 16 (the bare minimum for testing a SINGLE interaction effect).
Devin Incerti retweeted
“This article will review fundamental conceptual problems with the prediction of outcome risk and heterogeneity of treatment effect (HTE), as well as the limitations of conventional (one-variable-at-a-time) subgroup analysis.” bmj.com/content/363/bmj.k424…
Devin Incerti retweeted
Clinicians: a topic that comes up repeatedly is the false belief that no outcome estimates apply to individual patients. Best recipe: covariate-adjusted treatment effect from an RCT coupled with a risk model from the RCT or a good cohort: fharrell.com/post/ehrs-rcts
Devin Incerti retweeted
Landmark review paper on Bayesian model averaging in economics by Mark Steel: aeaweb.org/articles?id=10.12… or earlier free version: arxiv.org/pdf/1709.08221.pdf. A impressive paper, combining deep scholarship with well supported opinion.
Devin Incerti retweeted
I came across this interesting article about Causality in Machine Learning
blog.fastforwardlabs.com/new… which also links to a recent panel on causality youtube.com/watch?v=ynVr_zzU… It is interesting to see how young ML researchers react to the new hype. Very encouraging #Bookofwhy.
Devin Incerti retweeted
Original model contains errors
Article is published
Model is adapted to another setting
Another article is published
Errors in model are finally identified
Second article is retracted ncbi.nlm.nih.gov/pubmed/3013…
WE NEED GREATER #TRANSPARENCY EARLIER IN THE #MODELLING PROCESS
Devin Incerti retweeted
Came across 'demotivate', an interesting Stata command .
So I just "had to" port it to R on the commute home yesterday. Find it in package dang on @GitHub.
#rstats #andIdontactuallyeverusedstatamyself
/cc @FelipeOrjuelaR @KevinDenny


























