What is the 'Barkson paradox' that produces counterintuitive statistical results such as 'good-looking guys are unpleasant'?



The experience you have gained can help you make decisions, but sometimes that experience can distort your decisions. Professor

Lionel Page of the Faculty of Economics at the University of Technology Sydney explains the 'Barkson paradox, ' a type of bias that distorts human judgment.

Burkson's paradox is that 'the outward correlation can be different from the true correlation.' As an example, Professor Page cites the proposition that women say, 'I don't like good-looking guys.'




Regarding this proposition, let's consider the correlation between hot (degree of handsomeness) and nice (degree of goodness of personality) on the premise that 'personality and appearance are irrelevant'. In this premise, personality and appearance are irrelevant, so 'good-looking guy but unpleasant (upper left)' 'good-looking guy (upper right)' 'busy but nice person (lower right)' 'busy but unpleasant guy (lower left)' Is considered to have a similar distribution.



However, for women, 'a ugly and unpleasant guy (lower left)', 'a little handsome but pretty unpleasant guy (a part of the upper left)', and 'a pretty ugly and a little nice person (a part of the lower right)' are dates. Because it is not the target of, it will be regarded as 'not done'. If only the remaining dating men are analyzed, the correlation between hot and nice becomes a blue line, and it is said that 'hot and nice have a negative correlation (= good-looking guys tend to be unpleasant)'. , A conclusion different from the premise is drawn. Here is an example of Berkson's paradox.



Another typical example of Berkson's paradox is that 'when considering new college students, there is a negative correlation between common first- stage (SAT) scores and high school academic performance (GPA) scores.' Of course, this is counterintuitive, as those with better academic performance in high school are considered to have higher college entrance exam scores.




However, in university entrance exams, there are two phenomena: 'People with too good SAT and GPA apply for a higher level university (upper right)' and 'If SAT and GPA are too low, they will be dropped by the entrance exam (lower left)'. Will occur. Therefore, if you exclude these two, the blue part of the graph remains, and if you analyze only this part, a negative correlation will be established between the SAT and GPA.



The proposition 'A terrible live-action movie is born from a good novel' is contrary to the intuition that 'A good novel should make a good live-action movie', but it can be explained by Burkson's paradox. When asked, 'Think of a live-action movie based on a novel,' people usually think of 'a good live-action movie' or 'a live-action movie based on a good novel,' and 'a bad live-action movie of a bad novel.' It is not. If the scope of analysis is limited to live-action movies, which are often remembered in this way, there will be a negative correlation between the goodness of the original novel and the goodness of the movie.




Burkson's paradox is a frequent issue in the fields of data science , such as medical statistics and biostatistics, and is considered an issue to keep in mind when dealing with statistics.

in Note, Posted by darkhorse_log