Sabitlenmiş Tweet
Your inability to think statistically is what makes you lurch from unjustified premise to illogical conclusion via irrelevant data.
English
Stephen John Senn
47.2K posts
Stephen John Senn
@stephensenn
CH & UK stats imposter, skier & Munroist: disappointing, humourless, confusing or disingenuous. DOI https://t.co/JN2UJQW1Lg
Edinburgh, Scotland Katılım Eylül 2009
1.1K Takip Edilen12K Takipçiler
@duolingo you won’t let me leave the screen I did not ask you to present me until I commit to a streak with a friend. I am a paying customer. Stop bullying me.
English
Farewell, Adieu
I'm off where skies are blue
I hope to see you too
English
@EE Thanks. I will try it. Your welcome to Switzerland message gave me the same number but ending 250 not 150.
English
@EE I am in Switzerland, I can't log on because the passcodes you claim you have sent me aren't getting through and the number you have given me to call you doesn't work. What do you suggest?
English
@AdanBecerraPhD @RWJE_BA @5_utr @f2harrell It’s kind of difficult to explain to a community that a) is only interested in point estimates because b) what only matters is what happens when “n” goes to infinity but c) can’t explain what “n” is.
English
@AvantiWestCoast any chance you could nudge your ticket inspector on the train that was due to arrive at Edinburgh at 22:21 but was delayed at Wolverhampton to give the passengers an update? Just a thought.
English
@FamedCelebrity @Forms_Respecter His statue is not far from where I live and his hat is frequently more colourful images.app.goo.gl/JTiT7N4uhWZy1n…
English
@Forms_Respecter Maybe it was a shower cap.
English
David Hume's Anti-Miracle Argument Is Poor
Hume didn’t believe in miracles. But then Hume didn’t believe we could know any cause.
He was shy about publishing his anti-miracle argument; they say it was because of fear of Christians. I think it was because he knew it was poor.🧵
English
@yudapearl @f2harrell My opinion is that a fusion of causal inference and statistics would be very powerful. DAGs as currently used don’t seem to tell us how to analyse complex experiments. Some of the causal work on observational data is very interesting.
English
Our weekly harvest of CI publications has swelled to 8 pages full of exciting titles: ucla.in/43qXI4x
Evidently, scientists are not impressed by this week's discovery by some by RCT gurus that "Causal Inference is the Scam of the Century"
@f2harrell
English
@soboleffspaces @yudapearl I like de Finetti “ The mathematician abstracts from reality, falls in love with the abstraction and then blamed reality for not conforming to it” (From memory.)
English
@yudapearl I like the parallels with confounding.
At the same time, Cardinal Bellarmine warns: enjoy your mathematical model, but don't irritate the consensus by suggesting the model represents reality.
English
Illuminating thread on Galileo's insufficient arguments. A tiny point to add on the analogy between Galileo's quarrel with the Church and our debate with statisticians. The former was a matter of simplicity (of theories), the latter can be settled mathematically -- there is no statistical test for confounding.
Michal J A Paszkiewicz@MichalYouDoing
A 🧵of Galileo's arguments for Heliocentrism,& why they didn't make a convincing proof: 1. Tides Galileo was convinced of Heliocentrism by his tide model. Galileo's Dialogo was in fact a modification and extension of his 1616 Dialogue on the Ebb and Flow of the Tides. 🧵1
English
@ProfHayward @elmir1omerovic @kaulcsmc @f2harrell @yudapearl @drjohnm @GreggWStone However, the recent study from Exeter published in the Lancet shows how modelling of observational data might improve use of drugs for type 2 diabetes already studied in RCTs. news.exeter.ac.uk/faculty-of-hea…
English
@ProfHayward @elmir1omerovic @kaulcsmc @f2harrell @yudapearl @drjohnm @GreggWStone Indeed. We use RCTs to decide whether OSs will become possibles. An OS can collect useful info on rare side effects an RCT cannot but will require care and tricks to interpret. tandfonline.com/doi/abs/10.108…
English
More evidence that observational studies are an important part of EBM?
If RCTs were the "only game in town," there would be more disagreement. Am I wrong?
DOI: 10.1002/14651858.MR000034.pub3.
@kaulcsmc @f2harrell @yudapearl @stephensenn @drjohnm @GreggWStone
English
@yudapearl @RWJE_BA “The reason is that the individual effects... are suspect of contamination by selection bias and placebo effects”
How do you deal with this problem in your observational studies?
English
@stephensenn @RWJE_BA Sure, but do you see any difficulty in analyzing P(Harm) given E[Y_t-Y_c] and E[P(Y|t)-P(Y|c)], instead of E[Y_t-Y_c] and P(Y|t), P(Y|c), as was done in our addendum.
English
Do RCTs give us all the causal information we need? No! Consider an RCT study in which the Average Treatment Effect is Zero. Can we tell if treatment has no effect on any individual or, perhaps, kills 10% and cures another? Let's hear what the RCT Zealots say.
NonsparseOncologist@5_utr
@f2harrell @biostatsfun @JohanDH2O @RWJE_BA @AdanBecerraPhD @dylanarmbruste3 @trumanfrancis @yudapearl @bobududu16 @elmir1omerovic @GreggWStone @xyu_shi @GiulioGrossi @kaulcsmc @stephensenn @drjohnm People draw valid causality from RCTs every day completely without do notation, that’s for sure
English
@soboleffspaces @yudapearl Of course, I also expect that potato farmers understand about clusters so I wonder when you are going to accept my challenge to explain them to your causal guru.
English
@soboleffspaces @yudapearl Respect! Anyone whose family has a potato connection is clearly an expert at finding roots.
English
@FamedCelebrity It's always interesting to learn what moralists think is acceptable behaviour by politicians.
English
Neocons on suicide watch. All those years, all that sweat, energy, and money spent trying to convince us that Putin is Satan incarnate have failed.
They will lash out. Sales of pinot grigio in DC are going to skyrocket.
English
@yudapearl @RWJE_BA Explain to me where you did take into account the study effect for your observational data.
Please clarify where the self-designated precise answer as regards this may be found.
All I see is that you implicitly assume it is zero.
English
Conversations with a denigrator.
Denigrator: Your math is useless because you
didn't take into account effect X
CI: Here it is, a precise answer including effect X
Denigrator: useless, because you didn't take into
account effect Y
CI: Here it is, a precise answer including effects X
and Y.
Denigrator: Useless, Useless, no matter...
English
@elonmusk I agree. If he was a puppet his nose would be enormous.
English
"Liverpool was a hugely cosmopolitan city" is a myth that won't die - it pops up again on Radio 4 this week.
Really the % of foreign-born residents in 1931 was quite low, esp for a port city:
Liv: 0.8%
Manchester: 1.2%
Leeds: 1.2%
London: 3.7%
England: 0.8%
(1931 census)
English
@yudapearl @RWJE_BA A Mueller and Pearl note, "the individual effects, P(yt) and P(yc), are suspect of contamination by selection bias and placebo effects." What they don't explain is how, if this is a problem that RCTs have to face, observational studies somehow escape.
English
@yudapearl @RWJE_BA 2/2) Modern work on using historical data has not ignored study effects but tried to find ways of estimating them. See, for example, pubmed.ncbi.nlm.nih.gov/25355546/
English
@soboleffspaces @yudapearl Indeed. But this reminds me of De Finetti's advice to petroleum exploration CEOs who did not like probabilistic forecasts: "don't drill dry wells".
"Only grow big potatoes," is very useful advice but for one snag.
English
@yudapearl Even Fisher wouldn't go so far as to assume that every potato from the plot with the higher average yield is bigger than the potatoes from the plot with the lower yield.
English