This video presents a nice example of the problem of false positives, especially when the baseline probability of having a disease is low. However, the comparison between the frequentist approach and the Bayesian approach is misguided here. Indeed, taking into account the baseline probability of Dengue into account is not "taking a Bayesian look", it is simply using Baye's theorem. A frequentist would compute the probability of the subject having the Dengue in the exact same way! The difference between frequentist and Bayesian statistics is much deeper than illustrated in this video.
I was thinking the same... He conflates or rather quickly jumps the gap between Bayes' theorem in the framework of probability theory, and Bayesian as a method of statistical inference.
Two points: 1. Until the twentieth century one could not have an appointment to a British school of higher learning unless you were ordained. No one says when mentioning Faraday or Maxwell, the most reverend. 2. Bayes was a member of the Royal Academy of Science, not so obscure to his peers.
It would help if you had included the definition of "False Positives," or the definition of "Specificity". The "Positive test results" are only 9.1% accurate, because the accurately diagnosed to have the disease results, "True Positives," are mixed with the inaccurately diagnosed people of having the disease, "False Positives". It might also be confusing for the viewer that why there is only one person with the disease, and why he is accurately diagnosed to have the disease? Imagine we have 10,000,000 people, which only 0.01% of them have the disease. And the test accuracy is 99.9%. Total people ---> 10,000,000 Have the disease ---> 1000 ---> D Don't have the disease ---> 9,999,000 ---> no-D D people: True positives: 999 of D will accurately be diagnosed positive. False negatives: only 1 person of that 1000 D will inaccurately be diagnosed as negative. no-D people: True negatives: from the 9,999,000 no-D, 99.9% or 9,989,001 will be accurately diagnosed negative. False positives: 9,999 of no-D people will inaccurately get a positive test result. Total true results = True positives + True negatives = 999 + 9,989,001 = 9,999,000 Total positives results = True positives + False Positives = 999 + 9,999 = 10,998 (True results) / (Total results) = 9,999,000 / 10,000,000 ------------------------------> 99.9% Test accuracy (True positives) / (Total positives) = 999 / 10,998 ------------------------------> 09.1% Positive results accuracy
i recently watched a vid about 'bayes theorem and some of the mysteries it has solved' or something. turing was one example, the vid mentioned that during the process, one of the men working with turing asked him 'isn't this bayes theorem?' to which he replied 'i suppose so,' which makes me think he didnt set about it with the name in mind
+FAFAWI2010 Though I didn't rate the video I would guess it is because it was a 5+ minute teaser trailer... for Bayes. That is, almost entirely devoid of actual information concerning 'A short history of probability'
Ok, I get it. The chances are getting it are 1 in 10 000 - but the chances of the test being wrong are 10 in 10 000. The test will give the 11 positive results, so there will be 9989 that get a negative result. Basically, you have a much higher chance of the test being wrong that actually having it. But I would definitely start panicking...
This video presents a nice example of the problem of false positives, especially when the baseline probability of having a disease is low. However, the comparison between the frequentist approach and the Bayesian approach is misguided here. Indeed, taking into account the baseline probability of Dengue into account is not "taking a Bayesian look", it is simply using Baye's theorem. A frequentist would compute the probability of the subject having the Dengue in the exact same way! The difference between frequentist and Bayesian statistics is much deeper than illustrated in this video.
I was thinking that too.
i am currently struggling to understand bayes theorem, can you please suggest a video that you believe would be helpful? Thank you
Thanks you for saving me from writing a sternly worded response. ;-)
I was thinking the same... He conflates or rather quickly jumps the gap between Bayes' theorem in the framework of probability theory, and Bayesian as a method of statistical inference.
Well, the video is pretty short, therefore treats all the ideas it presents superficially.
Two points: 1. Until the twentieth century one could not have an appointment to a British school of higher learning unless you were ordained. No one says when mentioning Faraday or Maxwell, the most reverend. 2. Bayes was a member of the Royal Academy of Science, not so obscure to his peers.
i'm glad that circumstances lead me to this video.
Deserves more views......
It would help if you had included the definition of "False Positives," or the definition of "Specificity".
The "Positive test results" are only 9.1% accurate, because the accurately diagnosed to have the disease results, "True Positives," are mixed with the inaccurately diagnosed people of having the disease, "False Positives".
It might also be confusing for the viewer that why there is only one person with the disease, and why he is accurately diagnosed to have the disease?
Imagine we have 10,000,000 people, which only 0.01% of them have the disease.
And the test accuracy is 99.9%.
Total people ---> 10,000,000
Have the disease ---> 1000 ---> D
Don't have the disease ---> 9,999,000 ---> no-D
D people:
True positives: 999 of D will accurately be diagnosed positive.
False negatives: only 1 person of that 1000 D will inaccurately be diagnosed as negative.
no-D people:
True negatives: from the 9,999,000 no-D, 99.9% or 9,989,001 will be accurately diagnosed negative.
False positives: 9,999 of no-D people will inaccurately get a positive test result.
Total true results = True positives + True negatives = 999 + 9,989,001 = 9,999,000
Total positives results = True positives + False Positives = 999 + 9,999 = 10,998
(True results) / (Total results) = 9,999,000 / 10,000,000 ------------------------------> 99.9% Test accuracy
(True positives) / (Total positives) = 999 / 10,998 ------------------------------> 09.1% Positive results accuracy
The style of this video 🤌
Needs citation on how Alan Turing used Bayes Theorem to crack the Enigma. No evidence of such thing happening is present.
i recently watched a vid about 'bayes theorem and some of the mysteries it has solved' or something. turing was one example, the vid mentioned that during the process, one of the men working with turing asked him 'isn't this bayes theorem?' to which he replied 'i suppose so,' which makes me think he didnt set about it with the name in mind
why the dislikes !!! people are a meystry
+FAFAWI2010 Though I didn't rate the video I would guess it is because it was a 5+ minute teaser trailer... for Bayes. That is, almost entirely devoid of actual information concerning 'A short history of probability'
+FAFAWI2010 Completely biased.
He didnt explain how bayes theory was used in AI
99.9% of 10,000 is 9990 not 9989, don't understand why you guys said 9989.
99.9% of 9.999 = 9909.001?
I guess the 1 guy who has it isn't included in the 10 inaccurate results. But I'm generally confused by all this :/
Ok, I get it. The chances are getting it are 1 in 10 000 - but the chances of the test being wrong are 10 in 10 000. The test will give the 11 positive results, so there will be 9989 that get a negative result. Basically, you have a much higher chance of the test being wrong that actually having it.
But I would definitely start panicking...
approximately 1 in 10 people are frequentists
You need far more data to be sure about that, Bayes!
SUPER GOOD VIDEO ❤️❤️❤️❤️😊
love this video
"Looking for free shwag" lol you mean weeeeeeed? X)
Yes fellow bot friends it do be working *wink wink
The probability for dislikes is roughly 1 in 10 for this video currently. ;^)
2:22 where has this 1% come from???????
He says "1 person", not percent.
@@PuhuTube yes, you're right don't know how I misheard that?!
Much easier to understand than 3B1B video.
FREE SHWAG
Idealism
this video is pretty baysed
bayzed
유희상
김은하
Ideologie
neva name a boat after a math fk...
김은하