STA305H1 Lecture Notes - Lecture 5: False Discovery Rate, Family-Wise Error Rate, Type I And Type Ii Errors
18 Jan 2018
School
Department
Course
Professor
Document Summary
A typical study in medical or social sciences may involve investigating the e ects of many predictor variables on a variety of outcome (or response) variables. P-hacking is a pejorative term used to describe the process whereby researchers focus on the results of hypothesis tests whose p-values are small. An example of a study where p-hacking was consciously used to produce misleading results is given at http://io9. com/i-fooled-millions-into-thinking-chocolate-helps-weight-1707251800. Suppose we want to test null hypotheses h (k) for k = 0 [ k(x1, , xn)] where the expected value above assumes that the null hypothesis h (k) is true. In isolation, if k(x1, , xn) = 1, we would reject the null hypothesis h (k) (at level ) in favour of. 0 its alternative h (k) are true then the number of false rejections of the null hypothesis (that is, type i errors) will be approximately m, which may be unacceptably large.
