PSY 370 Lecture Notes - Lecture 20: Item Response Theory, Snob, Psy
Document Summary
Discrimination in this context is a good thing! Like how a wine snob would have a discriminatory palate. We want items that distinguish between people who are high in the attribute and people who are low. If everybody scores the same, the item doesn"t tell us anything about the construct! Two statistics we can use to evaluate item discrimination: ***these give us the same information, even though the way we get to them differently. Rank-order test scores and divide into 3 groups (divide into thirds) Calculate the % of people in the high group who got the item right (u) and the % of people in the low group who got it right (l) U and l should be in the format of percentages or decimals. Good discrimination when d is high and positive. We are throwing information away that could be beneficial!! Does not adapt well to items that are scored dichotomously.