Lecture 10

Question 1

When testing multiple outcomes, what are the possible different outcomes that can occur? Which are random? Which cannot be observed?

Answer 1

Question 2

What is the FDR?

Answer 2

Question 3

What is the main goal of FDR?

Answer 3

Question 4

What is the Benjamin & Hochberg (1995) procedure?

Answer 4

Question 5

What is the Benjamini & Hochberg (1995) Theorem?

Answer 5

Question 6

What is the difference between Benjamin & Hochberg and Holms procedure?

Answer 6

Question 7

How do Benjamin & Hochberg vs. Holm compare with thresholds?

Answer 7

Question 8

What is the Bejamini & Yekutieli (2001) Theorem?

Answer 8

Question 9

What is the PRDS property?

Answer 9

Question 10

Why and by whom was the PRDS property introduced?

Answer 10

Question 11

What is the Benjamini & Yekutieli procedure for multiple testing?

Answer 11

Question 12

Will the Benjamini & Yekutieli procedure for multiple testing procedure have more power than the Benjamini & Hochberg (1995)
procedure?

Answer 12

The answer ’No, it does not but it has other advantages.’ is the right one because we just convinced ourselves that it does not have more power yet it provides FDR control regardless of the distribution of the test statistics (see next theorem on the
next slide).