P-Values, MacKinnon, and Sample Length

Why does this course record T next to every Gate 3 p-value in the fold log?

• Because the p-value formula needs T to be valid
• Because the critical value depends on T, so without it nobody can recompute the decision
• Because FDR divides each p-value by T

A residual statistic is −2.5. Someone says it passes because it is beyond −1.96. What is wrong?

• Nothing, −1.96 is the usual 5% cut-off
• −1.96 comes from the normal distribution; the residual test needs about −3.34 at 5%
• The cut-off should be +1.96 because the test is one-sided