Read on Twitter
A running list of errors of reasoning and evidence I see in discussions on sars-cov-2 policy
[a living thread]
[a living thread]
1. Basing your views on some, but not all, data on a topic or question, i.e. cherry picking
Please read well done meta-analyses!
Please read well done meta-analyses!
2. Faulting a study for its limitations, while basing your conclusion & view on studies with greater limitations
3. Inquiring excessively about the motivations/ agenda of folks who disagree with you, but never questioning motivations/ agenda if they agree with you
4. Debating the credentials of folks who disagree with you, while celebrating the amateur savant who agrees with you
5. Drawing conclusions about medical therapies based on non-randomized evidence
good luck
good luck
6. Finding fault with a study or presentation-- when that fault is irrational or excessive--
e.g. despite Recovery protocol, statistical analysis plan being posted online, folks wanted to wait till results were in peer review journal SMH
e.g. despite Recovery protocol, statistical analysis plan being posted online, folks wanted to wait till results were in peer review journal SMH
7. Sanctifying peer review - a process that is 1. hard on minority view points and 2. a porous filter for quality aka not that good
8. Still sanctifying peer review after SURGISPHERE
corollary: citing surgisphere after retraction
corollary: citing surgisphere after retraction
9. Analyzing risk factors for disease or bad outcomes that ignore the table 2 fallacy https://academic.oup.com/aje/article/177/4/292/147738
10. Publishing your articles in a journal that you are the editor of & press releasing it
11. Risk prediction models https://www.medrxiv.org/content/10.1101/2020.03.24.20041020v2
12. Considering only the point estimate and not the confidence interval
13. Considering the confidence interval but not the pre-test probability
14. Misinterpretation of subgroups
Lack of understanding of interaction coefficients/ prespecification
Lack of understanding of interaction coefficients/ prespecification
15. Comparing data that was collected differently/ coded differently
16. Comparing data that you have no idea how it was coded
17. Demonizing people who disagree with you
18. Running massive uncontrolled experimental studies AND complaining that there are not enough people to randomizing
Like eating a bag of m&ms and asking where they went
Like eating a bag of m&ms and asking where they went
19. Vaccine discussions/ considerations that analyze the sars-cov-2 vaccine as one analyzes the pneumovax --i.e. failing to awknowledge the massive differences in the situation
20. The US clinical trial apparatus
21. Marrying your political views to your scientific ones
22. Not being able to tell you have paired your political views with your scientific ones
23. Petitions - i.e. confusing garnering signatures with generating evidence for your worldview
24. Being persuaded by powerful or well delivered anecdotes that fail to consider the denominator, and fail to rise to the level of data (not the plural of anecdote)
25. When 2 countries have different covid outcomes attributing all those differences to choices under human control w/o careful studies that confirm all differences were due to human choices
26. Not pursuing some scientific questions/ dialogs for fear that the results may be misinterpreted
The head in sand fallacy
The head in sand fallacy
27. A few more fun ones in this discussion with @PaulSaxMD from the Harvard Medical School https://academic.oup.com/ofid/article/7/12/ofaa581/6028438
28. Saying something is a parachute that does not have a 99.9999% absolute risk reduction & is nothing at all like a parachute
http://cmajopen.ca/content/6/1/E31.long
http://cmajopen.ca/content/6/1/E31.long
29. Greedy reductionism
If a mechanism makes sense, the treatment works
If an intervention is good, any effort to do more is better (even if the effort has offsetting harms or backfires)
If a mechanism makes sense, the treatment works
If an intervention is good, any effort to do more is better (even if the effort has offsetting harms or backfires)
