Statistical fallacies

Question 1

We think that statistically independent events are related when they are not
e.g. likelihood of flipping heads if just got tails

Answer 1

The Gambler’s Fallacy

Question 2

Just because two variables vary together does not mean that one caused the other (no causal connection)

Answer 2

Mistaking correlation for causation fallacy

Question 3

Conjunction fallacy

Answer 3

Thinking that the conjunction of two events is more likely than a single general event
Higher probability of A occurring than A and B (as A*B < A, unless A = 1/is definite)

Question 4

Mistaking statistical significance for clinical significance

Answer 4

Finding statistical significance does not mean that the thing is clinically significant
A significant statistical difference between two variables means that it was likely not due to chance, not necessarily that the treatment was clinically effective.
e.g. 3% difference between two things may not have any significant effect clinically

Question 5

Base rate fallacy (not common, very specific)

Answer 5

If presented with base rate information (general information) and information about a specific case, we tend to ignore the general information and only focus on the specific case.

Question 6

Truth inflation (not fallacy)

Answer 6

Made up of
File drawer effect: researchers who don’t find any effect just put their work away and never submit it for publication
Publication bias: research reporting that there is some effect is far more likely to be published than research that shows there is no effect (as more interesting)

Thus, much higher probability that a statistically significant event reported in a published study was just due to random chance (not seeing negatives, resulting in lots of false positives).

Question 7

Regression to the mean (not fallacy)

Answer 7

outliers are more likely to come back/regress to the mean because it is more likely to gather around the mean as it is the most common (and that mean will move closer to the outlier)