S- T- A- T205-ndash;- Elementary- Statistics-for-the- Biological-and- Life-课件.pptVIP

Example 11.9 – Rejection Region If instead we wanted to use a rejection region, then we reject k Ho: μ1 = μ2 = μ3 at a = 0.05 if Fs ≥ F.05(2,9) = 4.26. Since n Fs = 0.77 F.05(2,9) = 4.26, we once again o fail to reject Ho and conclude that no significant difference exists among diets. from Table 10 * Relation to t-dist’n Note that if I = 2, this ANOVA method will collapse precisely into the two-sample t-test: the (2-sided) inferences from the two methods will be identical in fact but, we cannot perform 1-sided tests with the F-statistic as we can with ts See Secs. 11.5–11.8 for more on ANOVA. * Multiple Group Model sample sample Observations from: # obsv’ns mean variance Y1 ~ N(μ1,s2) n1 Y2 ~ N(μ2,s2) n2 Y3 ~ N(μ3,s2) n3 M M YI ~ N(μI,s2) nI Notice: “dot” notation * Multiple Comparisons It’s natural to ask: why not now just compare all possible pairs μ1 – μ2, μ1 – μ3, μ2 – μ3, etc., each with a t-test (or conf. interval)? DEF’N: The PROBLEM of MULTIPLE COMPARISONS occurs when the same data set is used to make multiple inferences on I 1 associated parameters. * Error Inflation Suppose I = 3. Perform a t-test of Ho:μ1 = μ2 with false-positive error rate set to a = .05. Then, perform another test, now on Ho:μ1 = μ3. Jointly, the probability of making a false positive error is now larger than 5%!! In fact, it gets worse as I increases.Table 11.2 gives an illustration ? * Table 11.2 * Combined Analysis Error considerations make it more efficient to study an overall null hypothesis: Ho:μ1 = μ2 = … = μI (vs. a non-directional alternative HA: some difference among μi’s). Indeed, a more complete analysis seeks to incorporate information among all groups: e.g., we can estimate the common s2 using variation from all I groups. * ANOVA DEF’N: The ANALYSIS OF VARIANCE (ANOVA) among I 1 populations is used to make inferences about the I population means. The various components for an ANOVA are assembled from the data. We sta