You apply Bonferroni to 20,000 genes with an overall budget of 0.05. Which p values count as hits?

• Only p below 0.05
• Only p below 0.0000025
• Only p below 0.5

You run 20,000 gene tests, each with a 5 percent false alarm rate. About how many chance false alarms do you expect?

• About 1,000
• About 5
• About 20,000

A q value of 0.05 means about 5 percent of the whole gene list at that cutoff are expected false discoveries.

Circle one:   True   False

Scanning thousands of genes, why is false discovery rate control often more useful than Bonferroni?

• It runs faster on computers
• It needs no p values at all
• Guarding against any single false alarm buries real signals

What does Bonferroni-style family wise error control cap?

• The chance of even one false alarm across the whole list
• The fraction of false discoveries within the hit list
• The number of genes you may test

You run 10,000 tests, each with a 5 percent false alarm rate. Chance false alarms are about the rate times the test count. Type that number.

Answer: ______________

A student reads q equals 0.05 as a 5 percent chance that this one gene is a false alarm. What is wrong?

• Nothing: that is exactly what q means
• q speaks about the whole list, not about one gene
• q values only apply below 0.01

Pre filtering drops untrustworthy genes before testing. Why does this raise discoveries?

• It changes the meaning of the q value
• It shrinks the test count so real signals clear the bar
• It removes the need for any control