ANOVA,t,andPower-EastCarolinaUniversity.docVIP

East Carolina UniversityDepartment of Psychology Half-Tailed Tests in ANOVA One of my colleagues, who teaches Experimental Psychology, posed the following questions regarding research her students were doing. You may find my responses informative. From: ONeal, Margaret F ONEALM@MAIL.ECU.EDU To: Wuensch, Karl L WUENSCHK@MAIL.ECU.EDU Subject: power and significance Date: Saturday, November 17, 2001 11:24 AM Hi, Karl. Questions for you if you have a chance: Some of my 2210 students used the SPSS program to analyze the data fortheir experiments. Initially, they used the t-test, but some had proposed a directional hypothesis, so they halved the significance level reported by SPSS. After discussing power in lecture, I asked them to run their analysis again using SPSS ANOVA in order to generate power and magnitude of effect output. 1. Though the F test is one-tailed, it yields the same significance level as the two-tailed t-test. Why is that? 2. And how does it affect the accuracy of the observed power value SPSS yields, in that a one-tailed t-test is more powerful than a two-tailed? One wouldnt double the power reported. Thanks for sharing your wisdom if you have time, Maggie Karl’s Response ??? The ANOVA F is one-tailed, but it is not directional.? As the difference between sample means grows larger (in EITHER direction), the F grows larger.? If the null is true, the expected value of F is about 1, since the numerator (MSbetween groups)?estimates error and any treatment effect, and the denominator (MSwithin groups) estimates only error.? In other words, the ANOVA F appropriately employs a one-tailed p to test nondirectional hypotheses.? If there is only one df in the numerator, the ANOVA F will be exactly equal to the square of the t computed on the same means and the Fs one-tailed p will be absolutely equivalent to the ts two-tailed p.? So what if you want to test a directional hypothesis with ANOVA F?? Just apply the multiplication rule