PSY 350 Lecture 38: PSY350 3-24-18
Document Summary
When we do have theory to indicate that a particular pair of means should differ, we can test this with a planned contrast. We create a contrast coefficient (sometimes called y) and test whether it is equal to zero. Lots of ways we can set up these contrasts, but the sum of the weights must equal zero. If so, planned contrasts give you more power than post hoc tests. You"re not testing everything, so your risk of type i error is lower. Only use planned contrasts if your overall (omnibus) f test is significant. Otherwise, there is a good chance you are making a type i error. Just not more than k 1 of them. Contrasts get more complex with more elaborate study designs. If you don"t have a theory, you should use post hoc tests instead (and temper your conclusions accordingly) Report the necessary information when writing about the results of anova.
