PSYC 301 Lecture Notes - Lecture 17: Analysis Of Variance, Null Hypothesis, Family-Wise Error Rate
Document Summary
Anova: used for three or more groups/samples (eg. control group, drug group, therapy group, groups= different levels of the same iv. Factorial anova: examines 2 or more independent variables at once (each iv is called a factor; each group is a level, eg. Anxiety levels have an effect on verbal tasks/ pressure condition has an effect on verbal tasks. Interaction = does the effect of one factor depend on the level of the other factors: eg. Problems with dichotomization: different medians in different samples -> problems comparing across studies, lose effect size, lose power, losing es and power is like throwing out a third of the sample, yield false significant main effects and interactions. If you have a continuous variable, regression is better because don"t want to split a continuous variable into discrete categories: no problem using regression for categorical though, so anything you can do with. Omnibus null hypothesis: tests a wide variety of interactions, eg.