@ehljama.bsky.social
Reposted
An issue that often comes up is how to set a “non-inferiority margin” in a non-inferiority trial and how to ensure it is appropriate 1/8
#MethodologyMonday #109
#MethodologyMonday #109
January 27, 2025 at 7:32 AM
An issue that often comes up is how to set a “non-inferiority margin” in a non-inferiority trial and how to ensure it is appropriate 1/8
#MethodologyMonday #109
#MethodologyMonday #109
Reposted
"Public health officials and researchers need to acknowledge when they get things wrong."
"'Otherwise, we’re just going to get into a worse and worse cycle,'" our Ivan Oransky told @usatoday.com.
"'Otherwise, we’re just going to get into a worse and worse cycle,'" our Ivan Oransky told @usatoday.com.
Study retracted years after it set off an infamous COVID-19 treatment scandal
In 2020, then-President Donald Trump said he had been taking hydroxychloroquine to prevent a COVID-19 infection despite warnings over side effects.
www.usatoday.com
December 18, 2024 at 3:00 AM
"Public health officials and researchers need to acknowledge when they get things wrong."
"'Otherwise, we’re just going to get into a worse and worse cycle,'" our Ivan Oransky told @usatoday.com.
"'Otherwise, we’re just going to get into a worse and worse cycle,'" our Ivan Oransky told @usatoday.com.
Reposted
I updated fharrell.com/post/medml which has some related links #MachineLearning #ML #StatsSky #Statistics
Is Medicine Mesmerized by Machine Learning? – Statistical Thinking
Deep learning and other forms of machine learning are getting a lot of press in medicine. The reality doesn’t match the hype, and interpretable statistical models still have a lot to offer.
fharrell.com
December 16, 2024 at 6:16 PM
I updated fharrell.com/post/medml which has some related links #MachineLearning #ML #StatsSky #Statistics
Reposted
Wow that paper is delicious. If #MachineLearning can predict from an image of the knee whether the person has eaten refried beans in the past 12 months, overfitting/overinterpretation/declaring noise as signal must be in play. Predicting weird outcomes is a great way to do bullshit detection.
Of course they do. Statistical models can work this way too. The algorithms, in this sense, are not the problem. The problems lie in the foolish ways some try to apply them, and a broad lack of interest in actually evaluating whether the predictions they generate are clinically useful.
The authors point out that AI models base their predictions on sneaky shortcut effects all the time; they're just easier to identify when the conclusions (beer drinking) are clearly spurious.
www.nature.com/articles/s41...
www.nature.com/articles/s41...
December 16, 2024 at 3:00 PM
Wow that paper is delicious. If #MachineLearning can predict from an image of the knee whether the person has eaten refried beans in the past 12 months, overfitting/overinterpretation/declaring noise as signal must be in play. Predicting weird outcomes is a great way to do bullshit detection.
See my Tweet on the uses (few) and misuses (many) of MarketScan for observational medical research
x.com/ehlJAMA/stat...
x.com/ehlJAMA/stat...
x.com
x.com
December 16, 2024 at 10:58 PM
See my Tweet on the uses (few) and misuses (many) of MarketScan for observational medical research
x.com/ehlJAMA/stat...
x.com/ehlJAMA/stat...
Speaking of pharma advertising-Here is a fun fact
Pharma researches how many 'eyeballs' will view an ad before they run them. They collect extensive data on journal dissemination (copies mailed) and readership (via surveys).
For rest of this post: x.com/ehlJAMA/stat...
I
Pharma researches how many 'eyeballs' will view an ad before they run them. They collect extensive data on journal dissemination (copies mailed) and readership (via surveys).
For rest of this post: x.com/ehlJAMA/stat...
I
x.com
x.com
December 16, 2024 at 4:52 PM
Speaking of pharma advertising-Here is a fun fact
Pharma researches how many 'eyeballs' will view an ad before they run them. They collect extensive data on journal dissemination (copies mailed) and readership (via surveys).
For rest of this post: x.com/ehlJAMA/stat...
I
Pharma researches how many 'eyeballs' will view an ad before they run them. They collect extensive data on journal dissemination (copies mailed) and readership (via surveys).
For rest of this post: x.com/ehlJAMA/stat...
I
Reposted
It’s the run up to the holidays so time to resurrect #MethodologyMonday #19. Published c1906 in Nature by the statistician Francis Galton, here is the scientific way to cut your Christmas cake 🎂🎄🔪😊 Enjoy!
galton.org/essays/1900-...
galton.org/essays/1900-...
December 16, 2024 at 6:51 AM
It’s the run up to the holidays so time to resurrect #MethodologyMonday #19. Published c1906 in Nature by the statistician Francis Galton, here is the scientific way to cut your Christmas cake 🎂🎄🔪😊 Enjoy!
galton.org/essays/1900-...
galton.org/essays/1900-...
A Good Idea
Outlaw pharma advertising in TV.
I saw one ad last night for a cancer drug that costs $250,000 per year. This sort of thing should not be a subject for advertising directly to patients.
amgreatness.com/2024/12/16/a...
Outlaw pharma advertising in TV.
I saw one ad last night for a cancer drug that costs $250,000 per year. This sort of thing should not be a subject for advertising directly to patients.
amgreatness.com/2024/12/16/a...
An Open Letter to Robert F. Kennedy, Jr. › American Greatness
Dear Mr. Kennedy, Of the many issues you will tackle as Secretary of Health and Human Services, we implore you to work with FCC Commissioner Brenden Carr and immediately put an end to pharmaceutical…
amgreatness.com
December 16, 2024 at 4:13 PM
A Good Idea
Outlaw pharma advertising in TV.
I saw one ad last night for a cancer drug that costs $250,000 per year. This sort of thing should not be a subject for advertising directly to patients.
amgreatness.com/2024/12/16/a...
Outlaw pharma advertising in TV.
I saw one ad last night for a cancer drug that costs $250,000 per year. This sort of thing should not be a subject for advertising directly to patients.
amgreatness.com/2024/12/16/a...
Reposted
Glad you liked the blog article. Anyone doing feature exploration / predictive modeling when the sample size is not large enough to reliably analyze a single pre-specified feature is going to be disappointed when they try to validate the results. And #MachineLearning doesn’t help. #StatsSky
@f2harrell.bsky.social wrote a great blog post about this topic...one of my favorites
www.fharrell.com/post/marg/
www.fharrell.com/post/marg/
Unadjusted Odds Ratios are Conditional – Statistical Thinking
This article discusses issues with unadjusted effect ratios such as odds ratios and hazard ratios, showing a simple example of non-generalizability of unadjusted odds ratios.
www.fharrell.com
December 13, 2024 at 6:35 PM
Glad you liked the blog article. Anyone doing feature exploration / predictive modeling when the sample size is not large enough to reliably analyze a single pre-specified feature is going to be disappointed when they try to validate the results. And #MachineLearning doesn’t help. #StatsSky
Reposted
Cases of whooping cough in Michigan are at their highest level in at least 20 years, as vaccinations decline. The disease can be deadly to infants. Alarmed by the increase, doctors are pleading for families to vaccinate their children.
Whooping cough cases skyrocket in Michigan, as vaccinations decline | Bridge Michigan
A potentially deadly disease in infants, whooping cough — pertussis — has exploded in Michigan this year, in part because of decreased vaccination rates.
www.bridgemi.com
December 14, 2024 at 2:28 PM
Cases of whooping cough in Michigan are at their highest level in at least 20 years, as vaccinations decline. The disease can be deadly to infants. Alarmed by the increase, doctors are pleading for families to vaccinate their children.
Reposted
Calling bullshit - a skill that every applied statistician should master. Unfortunately many of the younger statisticians I’ve worked with sometimes lack the bravery to do so. The book looks like a must-have. #Statistics #StatsSky @carlbergstrom.com @carlzimmer.bsky.social
I wish this course on Calling Bullshit were offered everywhere.
Follow the link to find the syllabus, which is a thing of beauty. It looks like this will be available in a digital form for those of us not lucky enough to hear it in person.
Follow the link to find the syllabus, which is a thing of beauty. It looks like this will be available in a digital form for those of us not lucky enough to hear it in person.
As I mentioned three weeks ago, Jevin West and I are developing a sort of followup course to Calling Bullshit.
We've got the basic scrollytelling tech worked out and now need to choose a design.
We're not designers and would like to simply buy an HTML5 template if only for the colors and fonts.
We've got the basic scrollytelling tech worked out and now need to choose a design.
We're not designers and would like to simply buy an HTML5 template if only for the colors and fonts.
December 14, 2024 at 1:22 PM
Calling bullshit - a skill that every applied statistician should master. Unfortunately many of the younger statisticians I’ve worked with sometimes lack the bravery to do so. The book looks like a must-have. #Statistics #StatsSky @carlbergstrom.com @carlzimmer.bsky.social
The Gas Stove Health Effects Fallacy
See Twitter for full post about literature deconstructing the gas cooking appliance-adverse health effects literature-It is deeply flawed
x.com/ehlJAMA/stat...
See Twitter for full post about literature deconstructing the gas cooking appliance-adverse health effects literature-It is deeply flawed
x.com/ehlJAMA/stat...
December 14, 2024 at 2:41 PM
The Gas Stove Health Effects Fallacy
See Twitter for full post about literature deconstructing the gas cooking appliance-adverse health effects literature-It is deeply flawed
x.com/ehlJAMA/stat...
See Twitter for full post about literature deconstructing the gas cooking appliance-adverse health effects literature-It is deeply flawed
x.com/ehlJAMA/stat...
Reposted
Very interesting paper on confidence interval theory and misconceptions #StatsSky #Statistics
Prob this one doi.org/10.3758/s134...
The fallacy of placing confidence in confidence intervals - Psychonomic Bulletin & Review
Interval estimates – estimates of parameters that include an allowance for sampling uncertainty – have long been touted as a key component of statistical analyses. There are several kinds of interval ...
doi.org
December 4, 2024 at 3:54 PM
Very interesting paper on confidence interval theory and misconceptions #StatsSky #Statistics
Reposted
#Statistics and #RStats thought of the day: Before analyzing data or doing multiple imputation, explore missingness extents and patterns in your data: hbiostat.org/rflow/missing - see especially the interactive NA Combinations tab. #StatsSky
R Workflow - 6 Missing Data
hbiostat.org
November 24, 2024 at 12:49 PM
#Statistics and #RStats thought of the day: Before analyzing data or doing multiple imputation, explore missingness extents and patterns in your data: hbiostat.org/rflow/missing - see especially the interactive NA Combinations tab. #StatsSky
THE STATE OF EVIDENCE SUPPORTING SURGICAL PRACTICE
I recently completed my 3rd American Board of Surgery recertification examination. Kudos to the ABS for changing the examination process to one that makes far more sense than previous ones.
See x.com/ehlJAMA/stat... for the rest of this posting
I recently completed my 3rd American Board of Surgery recertification examination. Kudos to the ABS for changing the examination process to one that makes far more sense than previous ones.
See x.com/ehlJAMA/stat... for the rest of this posting
x.com
x.com
November 23, 2024 at 5:06 PM
THE STATE OF EVIDENCE SUPPORTING SURGICAL PRACTICE
I recently completed my 3rd American Board of Surgery recertification examination. Kudos to the ABS for changing the examination process to one that makes far more sense than previous ones.
See x.com/ehlJAMA/stat... for the rest of this posting
I recently completed my 3rd American Board of Surgery recertification examination. Kudos to the ABS for changing the examination process to one that makes far more sense than previous ones.
See x.com/ehlJAMA/stat... for the rest of this posting
Reposted
Just imagine a business that gets others to do all the work (the research and the peer reviews) then asks those who created the content with public funds to have to pay to make the content available to the public. Wow.
With the prices of publishing "Gold Open Access" these days maybe we should rename it "Platinum Exclusive PDF on the internet"
November 23, 2024 at 12:37 PM
Just imagine a business that gets others to do all the work (the research and the peer reviews) then asks those who created the content with public funds to have to pay to make the content available to the public. Wow.
Reposted
Even though this resulted in accidental release of data the public needs to know, this is another great example of why not to use Excel for scientific work …
November 16, 2024 at 1:15 PM
Even though this resulted in accidental release of data the public needs to know, this is another great example of why not to use Excel for scientific work …