ABSTRACT
When experiments involve the comparison of two sample means, the t test provides an appropriate test of significance. Some experiments, however, involve more than two samples. When the investigator tests the hypothesis that several sample means vary among themselves more than should be expected on the basis of random sampling, the t test is no longer satisfactory. Under these circumstances we need an omnibus test to determine the significance of variation among k sample means much as χ2 can be used to determine the significance of variation among k sample proportions. The procedure we use is called the analysis of variance (ANOVA). As with χ2, ANOVA can be used when k ≧ 2; although when k = 2 the t test, as described in the previous chapter, is a more conventional way to treat the data.
