Paradoxes in Data Science

Question 1

Base Rate Fallacy

Answer 1

P(A | B) != P(B | A)

Question 2

Misleading Accuracy

Answer 2

If we have a test which claims to be 99% accurate, which states that 1% of people have cancer, we can just return all negative cases. Therefore 99% will fall in line with the truth (the fact that 99% of people don’t have cancer) and will leave one to be inaccurate.

Question 3

What rates are measured for any test?

Answer 3

• Sensitivity -> correct acceptance of a positive result
• Specificity -> correct rejection of a negative result

Question 4

Simpson’s Paradox

Answer 4

A trend in different groups can reverse when these groups are combined.
Essentially, we could have a group A, and a group B, which individually say different things. But when combined to make group C, it can change things completely.

Question 5

Ecological Fallacy

Answer 5

Refers to wrong interpretations of statistical data when inferences abouts individuals are deduced from their group statistics.