PSYC 3031 Study Guide - Final Guide: Multiple Comparisons Problem, Linear Regression, Homoscedasticity

• School: York University
• Department: Psychology
• Course: PSYC 3031
• Professor: Alyssa Counsell

Document Summary

False discovery rate: fraction of statistically significant results that are actually false positives (related to type 1 error) Base rate fallacy: the misinterpretation that low p-values mean minimal statistical error. Problem with using pairwise t-tests when there are k>2 groups: Increases type 1 error rates (false positives) at m(alpha) Can compare means when we have k>2 groups. Uses one test at a type 1 error rate of alpha. Compares amount of variability between groups to amount of variability within groups. Bonferroni (worst) t-test that corrects for familywise error rate. Makes adjustments according to base rate (rather than the familywise error rate like the others) Compare ratio of highest variance over lowest variance (must be less than 2) 3x4=12 cells -> 2 ivs with 3 levels in the first iv, 4 levels in the second iv. Only works when cell means are representative of the population. Type 2: modified weighting method (more power when there is no interaction)