"For this example, assume that the statistical test fails (that is, there is a statistically significant difference between the projected risk scores in the sample of 200 and the subsample of 20)." . . . .