Analysis of variance(English statistics).ppt

Analysis of variance Department of Health statistics Introduction In previous modules you have learned how to compare two populations. That is, we have learned statistical methods for testing if there is any significant difference between the population means. While the procedure is useful, it is limited because it can only be used to compare two populations at a time. There are many situations, however, in which it is used to compare more than two populations. Examples Clinical psychologist may wish to compare three different therapies; A school psychologist may wish to compare four different study methods. One possible procedure for comparing multiple populations is to perform multiple t tests. Multiple t tests will create two problems: 1. Multiple t tests require a lot of work. That is, all possible number of comparisons will be ,where k is a number of populations(treatments) to be tested. It is therefore clear that a dull work increases as k increases. 2.Each t test has a type I error rate of α. However, over the series of multiple t tests, the overall type I error rate does not stay at α but grows with each new comparison. Analysis of variance(ANOVA) The ANOVA is a procedure for comparing multiple populations that avoids the inflation of type I error rates(and is less work than performing many t tests). The trick is that the ANOVA makes only one comparison to determine if any of the populations differ from the rest.The type I error rate remains exactly at α. Completely randomized design(One-way ANOVA) What is completely randomized design? The technique called One-way ANOVA is usually used to test the hypothesis that three or more treatments are equally effective. It is appropriate to the experimental design in which the treatments of interest are assigned completely at random to the subjects on which the measurement of outcome are to be made. This design is called a completely randomized design. What is completely randomized design? E