Stats for Strategy HOMEWORK 4 Solution (updated 5/26/2015) A. (a) 97.94% (b) Ronald Fisher B. Topic 6 Examples 1. (Example 1 Case 2) (a) P -value = 1 − 0.822021 = 0.177979 (b) 0.396569 (c) P -value > 0.396569 (d) P -value < 0.177979 2. (Example 1 Case 3) (a) s21 = 37 s22 = 30.5 s23 = 62.5 s2p = average(37, 30.5, 62.5) = 43.33 (b) F = (c) MSG 500 = = 11.54 MSE 43.33 The variation between sample means is 11.54 times as large as expected, if the null hypothesis is true. (d) 1 − 0.998398 = 0.001602 (e) (1) F = 11.54 (2) P -value = 0.002 (3) Almost! For the Notes Prof. Whitten used additional MINITAB graphing options to make small changes such as “jiggling” the points horizontally so that they’re not obscured by the sample means. 1 3. (a) MSG = 102.8 MSE = 538.9307692 F = MSG = 0.190748062 MSE (b) P -value = 0.828612 (c) F = 0.19, P -value = 0.829 C. Textbook 1. Pretest scores (before reading programs are applied) (a) 1. µB ( or µ1 ) = mean pretest score of all children taught with Basal µD ( or µ2 ) = . . . . . . DRTA µS ( or µ3 ) = . . . . . . Strat H0 : µ 1 = µ 2 = µ 3 HA : At least one mean differs from the others 2. From MINITAB: F = 1.13 , P -value = 0.329 3. Fail to Reject H0 since P -value = 0.329 > 0.05 = α 4. There is not sufficient evidence to show any difference in mean pretest scores between the three groups of children. (b) Yes. The difference in sample means seems small compared to the within-sample variation in each of the three groups. Individual Value Plot of Pretest Score vs Group 17.5 Pretest Score 15.0 12.5 10.0 7.5 5.0 Basal DRTA Group 2 Strat (c) Yes. The Tukey CI’s comparing mean scores all contain 0. Tukey Simultaneous 95% CIs Differences of Means for Pretest Score DRTA - Basal Strat - Basal Strat - DRTA -4 -3 -2 -1 0 1 2 If an interval does not contain zero, the corresponding means are significantly different. µD − µB = (−2.951, 1.406) (d) µS − µB = (−3.542, 0.815) µS − µD = (−2.770, 1.588) (e) Probably not but it doesn’t hurt to look at Tukey CI’s anyway, just to be sure. The ANOVA test has already indicated no statistically-significant differences. (f) Let µ = mean pretest score for all children. A 95% confidence interval for µ is (9.045, 10.530) points. Combining the three samples into one large sample makes sense because the F test shows no real differences between the three mean scores. In this case µ is the common mean pretest reading score for all children: µ = µ1 = µ2 = µ3 (g) Yes. 3 2. Reading comprehension scores (after reading programs are applied) (a) 1. µB = mean test score of all children taught with Basal method µD = . . . . . . DRTA µS = . . . . . . Strat H0 : µB = µD = µS HA : At least one mean differs from the others 2. From MINITAB: F = 4.48 , P -value = 0.015 3. Reject H0 since P -value = 0.015 < 0.05 = α 4. There is sufficient evidence to show that mean test scores differ between teaching methods. (b) It’s difficult to judge from the graph whether there are any real differences in the three population means. For that reason, we should rely on the P -value. Individual Value Plot of Comprehensive Reading vs Group 60 Comprehensive Reading 55 50 45 40 35 30 B D S Group (c) Yes. The F test (ANOVA test) has indicated significant differences in mean reading scores. (d) One of the Tukey CI’s shows a significant difference between the Basal and DRTA methods. A 95% confidence interval for (µD − µB ) is (1.12, 10.25) points so we are 95% confident that using DRTA rather than Basal will result in improved test scores of between 1.12 and 10.25 points, on average. Neither of the other two Tukey CI’s shows a statistically significant difference: • No clear difference between the Basal and Strat methods • No clear difference between the DRTA and Strat methods (e) With 95% confidence, the company can claim that its DRTA method is superior to Basal but cannot make a similar claim for Strat. 4 3. (Bookstores) (a) 1. µ1 = mean age of all Bookstore 1 customers (in years) µ2 = . . . . . . Bookstore 2 µ3 = . . . . . . Bookstore 3 µ4 = . . . . . . Bookstore 4 µ5 = . . . . . . Bookstore 5 H0 : µ1 = µ2 = µ3 = µ4 = µ5 HA : At least one mean differs from the others 2. F = 472.46 , P -value = 0.000 3. Reject H0 since P -value = 0 < 0.05 = α . 4. There is sufficient evidence to show that mean customer age differs between bookstores. (b) There are 7 significant comparisons and 3 non-significant comparisons. (c) There are three distinct clusters. (d) Cluster 1 = Bookstores 1, 4, 5 Cluster 2 = Bookstore 2 Cluster 3 = Bookstore 3 (e) If the clusters are defined as in (d), Cluster 1 has youngest customers. Cluster 2 has intermediate customers. Cluster 3 has oldest customers. (f) Bookstore customers at the mall tend to be younger than bookstore customers downtown. (g) Cluster 3 = Bookstore 3 has the oldest customers. A 95% confidence interval for µ3 is (53.894, 56.906) years. (h) Cluster 1 = Bookstores 1, 4, 5 has the youngest customers. Let µ represent the (common) mean age of customers at these three stores, all located at the mall. A 95% confidence interval for µ is (18.782, 20.418) years. 5 4. (More about clusters) (a) • Cluster 1 = Basil and Strat methods • Cluster 2 = DRTA and Strat methods (b) The clusters are not distinct since they contain a teaching method (Strat) in common. (c) Since both clusters contain Strat, neither cluster is shown to be clearly superior to the other cluster. 5. SE 14.3 (a) 1. µ0 µ1 µ3 µ5 µ7 = = = = = mean Vitamin C content in bread immediately after baking . . . . . . 1 day after baking . . . . . . 3 days . . . . . . 5 days . . . . . . 7 days H0 : µ0 = µ1 = µ3 = µ5 = µ7 HA : At least one mean differs from the others 2. F = 367.74 P -value = 0.000 3. Reject H0 since P -value = 0 < 0.05 = α 4. There is sufficient evidence to show that mean Vitamin C content in bread changes over time. (b) Tukey 95% CI’s for adjacent time points: • (µ1 − µ0 ) = (−12.05 , −1.45) • (µ3 − µ1 ) = (−25.46 , −14.86) • (µ5 − µ3 ) = (−14.68 , −4.08) • (µ7 − µ5 ) = (−9.39 , 1.20) There are significant differences among the first three pairs of adjacent time points, but not between the last pair. Interpret: We are 95% confident that average Vitamin C content in bread (in mg per 100g of flour) decreases by between • 1.45 and 12.05 mg immediately after baking to one day after baking. • 14.86 and 25.46 one day after baking to three days after baking. • 4.08 and 14.68 three days after baking to five days after baking. 6 (c) Mean loss of Vitamin C in bread begins immediately after baking and continues until five days after baking, at which time it appears to level off between five and seven days after baking. (d) There are 4 distinct clusters of adjacent time points: Cluster Cluster Cluster Cluster 1 2 3 4 = = = = Immediately after baking 1 day . . . 3 days . . . 5 days and 7 days . . . 6. SE 4.5 Vitamin A (a) 1. µ0 µ1 µ3 µ5 µ7 = = = = = mean Vitamin A content in bread immediately after baking . . . . . . 1 day after baking . . . . . . 3 days . . . . . . 5 days . . . . . . 7 days H0 : µ0 = µ1 = µ3 = µ5 = µ7 HA : At least one mean differs from the others 2. F = 12.09 P -value = 0.009 3. Reject H0 since P -value = 0.009 < 0.05 = α 4. There is sufficient evidence to show that mean Vitamin A content in bread changes over time. (b) Tukey 95% CI’s: • (µ1 − µ0 ): (−0.354 , 0.134) • (µ3 − µ1 ): (−0.274 , 0.214) • (µ5 − µ3 ): (−0.149 , 0.339) • (µ7 − µ5 ): (−0.584 , −0.096) There are no significant differences among three pairs but there is a significant difference between the last pair. Interpret: We are 95% confident that the average Vitamin A content in bread decreases by between 0.096 and 0.584 mg per 100g of flour from five days after baking to seven days after baking. 7 (c) There’s no detectable loss in average Vitamin A content in bread until five days after baking, but there is a loss between five and seven days after baking. (d) There are 2 distinct clusters of time points: Cluster 1 = Immediately, 1 day, 3 days, 5 days after baking Cluster 2 = 7 days after baking Vitamin E (a) 1. µ0 µ1 µ3 µ5 µ7 = = = = = mean Vitamin E content in bread immediately after baking . . . . . . 1 day after baking . . . . . . 3 days . . . . . . 5 days . . . . . . 7 days H0 : µ0 = µ1 = µ3 = µ5 = µ7 HA : At least one mean differs from the others 2. F = 0.69 P -value = 0.630 3. Fail to Reject H0 since P -value = 0.630 > 0.05 = α. 4. There is not sufficient evidence to show that mean Vitamin E content in bread changes over time. (b) Tukey CI’s: • (µ1 − µ0 ): (−8.13 , 6.43) • (µ3 − µ1 ): (−5.88 , 8.68) • (µ5 − µ3 ): (−6.78 , 7.78) • (µ7 − µ5 ): (−9.93 , 4.63) There are no significant differences between any of these pairs of means, which is consistent with the fact that we did not reject the ANOVA null hypothesis! (c) There is no apparent loss in Vitamin E in bread (on average) over the first seven days after baking. (d) There is a single cluster of time points: Cluster 1 = Immediately, 1 day, 3 days, 5 days, 7 days after baking 8 (Exercise 7 solution on next page) D. Answers to Additional Practice 1. (a) 2. (a) 3. (e) 9 7. (a) and (b) Vitamin C, A, E Loss in Bread Over Time Vitamin C in Bread Over Time 50 40 tn ten oC C 30 ni m tai V 20 10 immediate Vitamin C: F = 367.74 one day three days P -value = 0.000 five days seven days (Also circle Vitamin C clusters on graph.) Vitamin A in Bread Over Time 3.4 3.3 nte tn 3.2 oC A ni 3.1 am itV 3.0 2.9 immediate Vitamin A: F = 12.09 one day three days P -value = 0.009 five days seven days (Also circle Vitamin A clusters.) Vitamin E in Bread Over Time 98 97 tn et 96 no C E 95 in m at iV 94 93 92 immediate Vitamin E: F = 0.69 one day three days P -value = 0.630 five days seven days (Also circle Vitamin E clusters.) (c) As the F statistics decrease, the P -values increase while the numbers of clusters decrease. 10
© Copyright 2025