Anyone work within annoying firewalls at work? RStudio usually takes 3-4 minutes to open, and it can take 30-40 minutes to download all the CRAN hosted packages I need to start a project. On the same network with my personal machine, the whole process is over in around 90 seconds. #rstats
I bet this will be interpreted by some racists as proof that migration is out of control -- "they don't even put English on the signs anymore!!1"
I lost around 15kg between end of my PhD and starting new job (due to the magic of stress) and I am legit swimming in my old clothes. I spent a few weeks looking like a street orphan fallen on hard times. 4 extra holes in my belt 😂
In fact the whole paper is hilarious
Excellent trial name
I have two university affiliations giving me access, and it still usually takes 3-4 minutes per paper navigating failed attempts or freezing browser windows. It would be funny if it wasn't a big chunk of my day-to-day
I just made this point on Gelman's blog. If you were an intelligence agency, you'd be following those closely. Just wait for the inside traders to bet big and you have plenty of advanced warning.
I am not sure I even have the mental energy to read fiction at the moment. A shame as I do miss reading for pleasure. Reading papers is draining.
Nice rec, I'll give them a follow
I can think of no surer way to make your spouse see the error of their ways than to point out how their current thinking is rife with confounders. Draw them a DAG while you're at it.
I don't actually know if any do. I guess I just don't read the media for science anymore! That said, I did have a complimentary subscription to New Scientist for a few months and enjoyed it a lot.
Probably because I'm 2/3 of the way through Rethinking and it's a neat phrase 😎
I've met so many US doctors who are nihilistic about saving patients only for them to get shot. Many have recounted to me that any attempts for them to confront this in the political sphere are met with the medical equivalent of "stick to sports"
Many such cases. I bet they think open source means every time you call a function, it's a Slavic teen porting into your PC to poke through your data.
My gut feeling is that it's because the whole process is backwards. Read the literature -> formulate your arguments -> cite those papers has been replaced by vibes -> find title that sounds vaguely relevant -> cite it. The latter is rapidly accelerated by ChatGPT recs (which are always wrong!)
Max entropy is the statistical version of "Measure twice, cut once." Multiverse analysis in practice is "Cut a bunch of times, and pick the ones you like the best."
This happens to me and all my colleagues constantly. It is clear to me that most citations only exist to give the veneer of scientific validity; nobody reads them and they serve no didactic purpose. You cite because it's something that "science" does.
I used to read the Guardian more regularly, but somehow it seems like they just can't write a good scientific article without being suckered in by stupid methods. Shame, used to be a good paper. Their football coverage is also rubbish these days.
MIMIC accounts for a huge proportion of all prediction modelling studies
I actually don't have a problem with utilise, but "leveraging AI" can fuck right off.
I was going to ask why you're dignifying them with a reply, but in fairness, this is pretty funny. You should say, "Of course a T A U R U S would say that."
I feel as though you know exactly why and are just being polite 😇
Weirdly I find I get pretty good advice from LLMs when I am quite explicit about the research question I want to answer. This is almost certainly a GIGO situation, and the GI is just...the body of existing research out there 🙃
Pretty sure it's the law that you have to leverage AI, you can't just use it.
I know this is a joke, but for anyone out there who might benefit: this was me when I first started my PhD, then my code failed to replicate on a published paper because a function's default behaviour changed, and I had no idea which package version I was even using to bring with
My suspicion is that they hire the service provider with the best legal team rather than the best engineers, which sort of explains everything
I've met a distressing number of educators and students who believe stats needs more, not less, point-and-click. They love Excel for this reason. Granted, I agree the true base R zealots are few, and unlikely to accost people in the street for using pipes.
To be fair to them, I guess the main thing they wanted was to pin the blame on a vendor when clearly inadequate digital infrastructure failed. Easier to spend $500m on a service contract than to hire a good engineer.
