RT @pash22: "Overestimating Outcome Rates: Statistical Estimation When Reliability Is Suboptimal" https://t.co/9UOq7y9gJR
RT @pash22: "Overestimating Outcome Rates: Statistical Estimation When Reliability Is Suboptimal" https://t.co/9UOq7y9gJR
"Overestimating Outcome Rates: Statistical Estimation When Reliability Is Suboptimal" https://t.co/9UOq7y9gJR
@pash22 @McGillOSS @crackedscience And, for those who want more statistical detail https://t.co/76YfDzEqGD
@drStuartGilmour @Nrg8000 @jnzst 1/ What is your theory of why "volatility" should be considered independent of reliability? One should not assume volatility is due to random variation when you can simply model it in a mixed model. https://t.co/jBbDwue5iI
@ESMDcan123 @statsepi @deaneckles Here's a paper abt how random error can induce HUGE bias. Sry, it's long but I'm bad at short explications. https://t.co/jBbDwuvGHi
RT @ProfHayward: @traependergrast And here's a methods paper that I hope is a good explainer on the math error. https://t.co/a0jvKxXYkN
@traependergrast And here's a methods paper that I hope is a good explainer on the math error. https://t.co/a0jvKxXYkN
@mloxton @gorskon Been there, done that. The numbers don't die b/c most ppl prefer simple lies to dealing w/ complex, multifaceted problems. https://t.co/jBbDwue5iI https://t.co/r9LehOG5X5
@gorskon Those numbers are pure math & logic errors. https://t.co/jBbDwue5iI https://t.co/r9LehOG5X5
RT @anish_koka: Nice example of tribalism here. Where was the put-down when @SenSanders used this ‘evidence’ to forward his pet agenda ?…
RT @ProfHayward: Just FYI @nratv & @DLoesch, there is absolutely no evidence that malpractice is a common very cause of death in U.S. House…
RT @ProfHayward: Just FYI @nratv & @DLoesch, there is absolutely no evidence that malpractice is a common very cause of death in U.S. House…
Nice example of tribalism here. Where was the put-down when @SenSanders used this ‘evidence’ to forward his pet agenda ? Perhaps I missed it? https://t.co/jGrrjccEAH
RT @ProfHayward: Just FYI @nratv & @DLoesch, there is absolutely no evidence that malpractice is a common very cause of death in U.S. House…
Just FYI @nratv & @DLoesch, there is absolutely no evidence that malpractice is a common very cause of death in U.S. Household. /1 See: https://t.co/jBbDwuvGHi
@drjohnm @RogueRad In this f/u paper we tried to make the phenomenon of "shrinkage" (impact of suboptimal reliability) more accessible. https://t.co/jBbDwuvGHi
RT @ProfHayward: @MikeRoseMDMPH @RogueRad Thanks. Here is a paper in which we try to explicate this generalizable measurement phenomenon.…
@MikeRoseMDMPH @RogueRad Thanks. Here is a paper in which we try to explicate this generalizable measurement phenomenon. https://t.co/jBbDwue5iI https://t.co/s8dIScIJVY
@derickrpeterson @stephensenn Here is an example, tho admittedly, with a much greater source of unaccounted for random error than is likely to be the case here. https://t.co/jBbDwuvGHi
RT @JeremySussman: @bnallamo @venkmurthy @iwashyna @thebyrdlab @ProfDFrancis @iamritu @kdpsinghlab An obvious @ProfHayward classic. https:/…
@bnallamo @venkmurthy @iwashyna @thebyrdlab @ProfDFrancis @iamritu @kdpsinghlab An obvious @ProfHayward classic. https://t.co/RyyhIyxVe3
The statistical analysis in these studies can also frequently OVERestimate the number of preventable deaths from medical error. https://t.co/2PFBaHWA7Y
Also, when studies have tried to estimate correlation and causation for adverse events and death from medical error, there has been a wide range of inter-observer variability. https://t.co/2PFBaHWA7Y
@beyerstein @MarkHoofnagle @MWFriedberg @MaryDixonWoods @RogueRad @mikejohansen2 @venkmurthy @Skepticscalpel @theNAMedicine Detailed explainer attached, but here is simplest way I know2explain it: All tests, incling chart review of preventable deaths, have
Overestimating Outcome Rates: Statistical Estimation When Reliability Is Suboptimal https://t.co/P42jg0bIlP
Overestimating Outcome Rates: Statistical Estimation When Reliability Is Suboptimal https://t.co/V3voVpqw34 via @ProfHayward
@MWFriedberg @MaryDixonWoods @RogueRad @mikejohansen2 @MarkHoofnagle @venkmurthy @Skepticscalpel @theNAMedicine I fault NAMs (then IOM) post-publication response. Poor understanding of impacts of random measurement error was not well understood at the time
@MarkHoofnagle @RogueRad @mikejohansen2 @venkmurthy @Skepticscalpel Yes, adjusting for "avoidability" helps, but adjusting for reliability is also essential, & I have not seen that done (tho I could have missed it). The issue is explained in this paper
RT @ProfHayward: @medevidenceblog @iwashyna @geoghep This paper explains in greater detail the nature of the large bias introduced by ignor…
@medevidenceblog @iwashyna @geoghep This paper explains in greater detail the nature of the large bias introduced by ignoring random error in outcome assessment. It is a very common methodological error. https://t.co/jBbDwuvGHi
RT @ProfHayward: Math error after math error. 👇👇👇👇👇👇 https://t.co/jBbDwuvGHi https://t.co/GALXbm0W7F
RT @ProfHayward: Math error after math error. 👇👇👇👇👇👇 https://t.co/jBbDwuvGHi https://t.co/GALXbm0W7F
RT @ProfHayward: Math error after math error. 👇👇👇👇👇👇 https://t.co/jBbDwuvGHi https://t.co/GALXbm0W7F
RT @ProfHayward: Math error after math error. 👇👇👇👇👇👇 https://t.co/jBbDwuvGHi https://t.co/GALXbm0W7F
RT @Skepticscalpel: The work @ProfHayward refers to is a paper he co-authored called "Overestimating Outcome Rates: Statistical Estimation…
RT @ProfHayward: Math error after math error. 👇👇👇👇👇👇 https://t.co/jBbDwuvGHi https://t.co/GALXbm0W7F
The work @ProfHayward refers to is a paper he co-authored called "Overestimating Outcome Rates: Statistical Estimation When Reliability Is Suboptimal" Full text here https://t.co/pFpd3VlJyG
RT @ProfHayward: Math error after math error. 👇👇👇👇👇👇 https://t.co/jBbDwuvGHi https://t.co/GALXbm0W7F
RT @ProfHayward: Math error after math error. 👇👇👇👇👇👇 https://t.co/jBbDwuvGHi https://t.co/GALXbm0W7F
RT @ProfHayward: Math error after math error. 👇👇👇👇👇👇 https://t.co/jBbDwuvGHi https://t.co/GALXbm0W7F
RT @ProfHayward: Math error after math error. 👇👇👇👇👇👇 https://t.co/jBbDwuvGHi https://t.co/GALXbm0W7F
RT @ProfHayward: Math error after math error. 👇👇👇👇👇👇 https://t.co/jBbDwuvGHi https://t.co/GALXbm0W7F
RT @ProfHayward: Math error after math error. 👇👇👇👇👇👇 https://t.co/jBbDwuvGHi https://t.co/GALXbm0W7F
Math error after math error. 👇👇👇👇👇👇 https://t.co/jBbDwuvGHi
@rwyeh Random error also biases variance estimates, and this can cause major problems w/ outcome rates as well. It's non-intuitive, but here is Tim Hofer & my attempt to explain this problem, which continues to be a common error. https://t.co/jBbDwuvG
For more discussion on the math error component see: https://t.co/jBbDwuvGHi https://t.co/Ay8TdiOYlt
RT @ProfHayward: @RogueRad @petermbenglish @Skepticscalpel The numbers R mainly due2MathErrors and misinfo. If the math had been done corre…
@RogueRad @petermbenglish @Skepticscalpel The numbers R mainly due2MathErrors and misinfo. If the math had been done correctly, the estimated # of deaths due to med errors would have been dramatically lower, & events were almost all underuse anyway, NO
Losing StatPower is inevitable when outcome has substantial random-error, but methods exist to adjustment for bias. https://t.co/jBbDwue5iI https://t.co/ovg9uhSV7s
@YFeyman @VinayPrasad82 The HSR is a tough read, but is a problem for all error measurement. https://t.co/xMUkLCfNh7 https://t.co/TS4akbhl5j