CHAPTER 12 ANALYSIS OF VARIANCE 1. The null and the alternative hypotheses are: H0: 1 = 2 = 3 H1: The population means are not all equal. We have selected = 0.05. We shall use as test statistic F = MST . MSE Since the population distributions are approximately normal with equal variance, the distribution of the test statistic, under H0, is F-distribution with df = (k-1, n-k) = (2, 9). This is an upper-tailed F-test. For df = (2,9), F = F0.05 = 4.26. The decision rule is: Reject H0 is favour of H1 if the computed F-value is greater than 4.26. For the given sample data: 86 4 = 5.083 12 7 9 3 3 3 4 x.1 = 8.25 ; x.2 = 3 ; x.3 =4 4 4 4 2 2 SSTotal = 8 5.083 4 5.083 74.92 . The overall mean is x SST = 4 8.25 5.083 4 3 5.083 4 4 5.083 62.1 . 2 2 2 SSE = SS Total – SST = 74.92 – 62.17 = 12.75. ANOVA Table Source of Variation Sum of Squares F Treatment Error Total Degrees of Freedom 2.17 12.75 74.92 2 9 11 Mean Square 31.085 1.4167 21.94 The computed F-value is 21.94, which is greater than the critical value of 4.26. So, there is sufficient evidence, at = 0.05, to reject H0, that is, to conclude that the population means corresponding to the three treatments are not all equal. 3. Let 1, 2, and 3 be the population means of family incomes in the area #1, area #2 and area #3, respectively. Then, the null and the alternative hypotheses are: H0: 1 = 2 = 3 H1: ’s are not all the same. We have selected = 0.05. 12-1 We shall use as test statistic F = MST . MSE Since the population distributions are approximately normal with equal variance, the distribution of the test statistic, under H0, is F-distribution with df = (k-1, n-k) = (2, 9). This is an upper-tailed F-test. For df = (2, 9), F = F0.05 = 4.26. The decision rule is: Reject H0 is favour of H1 if the computed F-value is greater than 4.26. For the given sample data: 64 68 78 = 71.25, 12 64 60 74 70 75 78 x.1 = 65.5, x.2 = 71, x.3 = 77.25 4 4 4 The overall mean is x SSTotal = 64 71.25 2 SST 78 71.25 = 364.25. 2 = 4 65.5 71.25 2 4 77.25 71.25 = 276.50. 2 SSE = SS Total – SST = 364.25 – 87.75 = 87.75. Thus, we get the following ANOVA table: Source of Variation Sum of Squares F Treatment Error Total Degrees of Freedom 276.50 87.75 364.25 2 9 11 Mean Square 138.25 9.75 14.18 The computed F-value is 14.18, which is greater than the critical value of 4.26. So, there is sufficient evidence, at = 0.05, to reject H0, that is, to conclude that the population means are not all equal. 5. Let 1, 2, and 3 be the population means corresponding to the three treatments. Then, the null and the alternative hypotheses are: H0: 1 = 2 = 3 H1: The population means are not all equal. We have selected = 0.05. We shall use as test statistic F = MST . MSE Since the population distributions are approximately normal with equal variance, the distribution of the test statistic, under H0, is F-distribution with df =( k-1, n-k ) = (2, 9). This is an upper-tailed F-test. For df = (2, 9), F = F0.05 = 4.26. The decision rule is: Reject H0 is favour of H1 if the computed F-value is greater than 4.26. For the given sample data: 12-2 The overall mean = x 64 68 12 78 3 2 The sample means are: x.1 8 10 = 9.667, x.2 3 SSTotal = (8.0 4.667) 2 = 4.667. = 2.2, x.3 3 4 5 4 2 4.0 4.667 = 116.67. SST = 3 9.667 4.667 SSE = SS Total – SST = 116.67 - 107.20 = 9.47. 2 = 4.0. 4 4.0 4.667 = 107.20. 2 ANOVA Table Source of Variation F Treatments Error Total Sum of Squares 107.20 9.47 116.67 Degrees of Freedom 2 9 11 Mean Square 53.600 1.052 50.96 The computed F-value is 50.96, which is greater than the critical value of 4.26. So, there is sufficient evidence, at = 0.05, to reject H0, that is, to conclude that the population means corresponding to the three treatments are not equal. Using Fisher’s LSD, we get a 95% confidence interval estimate for (1-2) as: ( x.1 x.2 ) t 2C MSE (1/ n1 1/ n2 ) (9.667 2.20) 2.9358 1.052(1/ 5 1/ 4) 7.467 2.02 (5.447,9.487) Since the interval does not contain 0, there is significant evidence to conclude that the population means of treatments 1 and 2 are different. 7. (a) Let 1, 2, 3 be the population means of number of month before a raise in salary was granted for CPA Inc., AB Intl., Acct ltd., and Pfisters, respectively. Then, the null and the alternative hypotheses are: H0: 1 = 2 = 3 H1: The population means are not all equal. We have selected = 0.05. We shall use as test statistic F = MST . MSE Since the population distributions are approximately normal with equal variance, the distribution of the test statistic, under H0, is F-distribution with df = (k-1,n-k) = (3,10). This is an upper-tailed F-test. For df = (3, 10), F = F0.05 = 3.71. The decision rule is: Reject H0 is favour of H1 if the computed F-value is greater than 3.71. 12-3 For the given sample data: The overall mean = x 14.3 18.1 16 The sample means are: x.1 x.3 12 18 12 4 3 16 12.0 ; x.2 13.2 14 15.33; x.4 12 10 4 11.5 ; 16 3 13.0. 14.0 16 13 78.0. 2 SSTotal = (12 13) 2 SST = 4 12.0 13.0 SSE = SS Total – SST = 78.0 - 32.33 = 45.67. 2 3 14.0 13.0 32.33. 2 Thus, we get the following ANOVA table: Source of Variation Sum of Squares Degrees of Freedom Mean Square F Treatments Error Total 3 10 13 10.78 4.567 2.36 32.33 45.67 78.00 The computed F-value is 2.36, which is less than the critical value of 3.71. So, we do not have sufficient evidence, at = 0.05, to reject H0, that is, to infer that the population means of number of months before a salary raise was granted of the four CPA firms are not all the same. (b) 9. Since the null hypothesis is not rejected in part (a), we do not need to perform the Tukey’s test. Let 1 and 2 be the population means corresponding to the treatments 1 and 2 respectively. Then, the null and the alternative hypotheses for treatments are: H0: 1 = 2 H1: 1 ≠ 2 We shall use as test statistic: F = MST MSE The population distributions are approximately normal with equal variance. Hence, under H0, the test statistic has approximately an F-distribution with df = (k-1, n-k-r+1) = (1, 2). This is an upper-tailed F-test. We have selected a significance level of 0.05. For df = (1, 2), F0.05 = 18.5. The decision rule is: reject H0 in favour of H1 if the computed F-value is greater than 18.5. 12-4 Let b1, b2, and b3 be the population means corresponding to the blocks A, B, and C, respectively. Then the null and the alternative hypotheses for blocks are: H0: b1 = b2 = b3 H1: the block means are not all equal. We shall use as test statistic: FB = MSB . MSE The population distributions are approximately normal with equal variance. Hence, under H0, the test statistic has approximately an F-distribution with df = (r-1, n-k-r+1 ) = (2, 2). This is an upper-tailed F-test. We have selected a significance level of 0.05. For df = (2, 2), F0.05 = 19.0. The decision rule is: reject H0 in favour of H1 if the computed F-value is greater than 19.0. The overall mean is x 46 37 6 The sample means are: x.1 46 44 = 42.3 ; x.2 35 31 = 36.5. 35 = 30.7; 3 3 46 31 37 26 44 35 x1. = 38.5 ; x2. = 31.5; x3. = 39.5. 2 2 2 2 2 SSTotal = 46 36.5 35 36.5 289.5. SST= 3 42.33 36.5 3 30.667 36.5 204.167. 2 2 SSB = 2 38.5 36.5 2 31.5 36.5 2 39.5 36.5 76. 2 2 2 SSE = SSTotal – SST – SSB = 289.5 – 204.167 – 76 = 9.333. We get the following ANOVA table: Source of Variation SS Treatment 204.167 Blocks 76.000 Error 9.333 Total 289.5000 df 1 2 2 5 MS 204.167 38.000 4.667 F 43.75 8.14 Our conclusions are as follows: Test for treatment means: 43.75 is greater than the critical value 18.5. Hence, there is sufficient evidence, at = 0.05, to reject H0, that is, to infer that the treatment means are not all the same. Test for block means: 8.14 is less than 19.0. Hence, we do not have sufficient evidence, at = 0.05, to reject H0, that is, to infer that the block means are not all the same. 12-5 11. Let 1, 2, and 3 be the population means of number of units produced per employee on day, afternoon and night shifts, respectively. Then the null and the alternative hypotheses are: H0: 1 = 2 = 3 H1: At least one of the ’s is different. MST . MSE We shall use as test statistic: F = The population distributions are approximately normal with equal variance. Hence, under H0, the test statistic has approximately an F-distribution with df = (k-1,n-k-r+1) = (2, 8). This is an upper-tailed F-test. We have selected a significance level of 0.05. For df = (2, 8), F0.05 = 4.46. The decision rule is: reject H0 in favour of H1 if the computed F-value is greater than 4.46. Let b1, b2, b3, b4, and b5 be the population means of number of units produced by Mehta, Lum, Clark, Kurz, and Morgan, respectively. Then the null and the alternative hypotheses for blocks are: H0: b1 = b2 = b3 =b4 = b5 ; H1: the means are not all equal. We shall use as test statistic: FB = MSB . MSE The population distributions are approximately normal with equal variance. Hence, under H0, the test statistic has approximately an F-distribution with df = (r-1, n-k-r+1) = (4, 8). This is an upper-tailed F-test. We have selected a significance level of 0.05. For df = (4, 8), F0.05 = 3.84. The decision rule is: reject H0 in favour of H1 if the computed F-value is greater than 3.84. 31 33 27 = 28.867. 15 31 28 25 26 35 27 x.1 = 30.0; x.2 = 26.0; x.3 = 30.6; 5 5 5 31 25 35 33 26 33 28 24 30 x1. = 30.3; x2. = 30.7; x3. =27.3; 3 3 3 30 29 28 28 26 27 x4. = 29.0; x5. = 27.0. 3 3 The overall mean is x SSTotal = 31 28.867 27 28.867 139.73 . 2 2 SST = 5 30.0 28.867 5 30.6 28.867 62.53 SSB = 3 30.3 28.867 3 27.0 28.867 33.73 SSE = SSTotal – SST – SSB = 139.73 – 62.53 – 33.73 = 43.47. 2 2 2 2 We get the following ANOVA table : Source of Variation SS Treatment Blocks Error Total df MS F 62.53 2 31.265 5.75 33.73 4 8.4325 1.55 43.47 8 12-6 5.4338 139.73 Our conclusions are as follows: Test for treatment means: 5.75 is greater than the critical value, 4.46. Hence, there is sufficient evidence, at = 0.05, to reject H0, that is, to infer that the mean production rates during different shifts are not all the same. Test for block means: 1.55 is less than 3.84. Hence, we do not have sufficient evidence, at = 0.05, to reject H0, that is, to infer that mean production rates of different employees are not all the same. 13. The null and the alternative hypotheses are: H0: 1 = 2 = 3 = 4 H1: Treatment means are not all equal. Since the population distributions are approximately normal, the distribution of the test statistic, F MST is approximately F-distribution, with df = (k – 1, n – k) = (4-1, 24-4) MSE = (3, 20). For df = (3, 20), F = F0.05 = 3.10. The decision rule is: reject H0 in favour of H1 if the computed F-value is greater 3.10. MSE = SSE SSE SSE . df ( for error ) n k 20 Hence, SSE = 20 (MSE) = 20(10) = 200. SST = SS total – SSE = 250 - 200 = 50. MST = F= SST 50 16.67. df ( for treatment ) 3 MST 16.67 1.667. MSE 10 This gives us the following ANOVA table. Source of Variation SS df MS Treatment Error Total 50 200 250 3 20 23 16.67 1.667 10 12-7 F Since the computed F-value (=1.667) is less than 3.10, we conclude that we do not have sufficient evidence, at = 0.05, to reject H0 in favour of H1, that is, to infer that the treatment means are not all equal. 15. Let 1, 2, and 3 be the population means of the prices of the toy at discount stores, variety stores and department stores, respectively. Then, the null and the alternative hypotheses are: H0: 1 = 2 = 3 H1: The population means are not all equal. We have selected = 0.05. We shall use as test statistic F = MST . MSE Since the population distributions are approximately normal with equal variance, the distribution of the test statistic, under H0, is F-distribution with df=(k-1,n-k) = (2, 12). This is an upper-tailed F-test. For df = (2, 12), F = F0.05 = 3.89. The decision rule is: Reject H0 is favour of H1 if the computed F-value is greater than 3.89. For the given sample data: The overall mean = x 23 5 15.867. 15 The sample means are: x.1 12 15 13.2 ; x.2 5 SSTotal = (12.0 15.867) 2 15 17 16.2 ; x.3 5 2 19 15.867 91.73. SST = 5 13.2 15.867 SSE = SS Total – SST = 91.73 - 63.33 = 28.40. 2 19 19 5 18.2. 5 18.2 15.867 63.33. 2 Thus, we get the following ANOVA table: SOURCE OF VARIATION SS Treatment Error Total 63.33 2 28.40 12 91.73 14 DF MS F 31.667 13.38 2.367 The computed F-value is 13.38, which is greater than the critical value of 3.89. So, there is sufficient evidence, at = 0.05, to reject H0, that is, to infer that the mean prices of the toy, in different types of stores, are not all equal. 12-8 17. Let 1, 2, 3, and 4 be the population means of number of crimes committed per day in Rec Centre, Key Street, Monclova and Whitehouse, respectively. The null and the alternative hypotheses are: H0: 1 = 2 = 3 = 4 H1: The population means are not all equal. We have selected = 0.05. MST . MSE We shall use as test statistic F = Since the population distributions are approximately normal with equal variance, the distribution of the test statistic, under H0, is F-distribution with df = (k-1, n-k) = (3, 20). This is an upper-tailed F-test. For df = (3, 20), F = F0.05 = 3.10. The decision rule is: Reject H0 is favour of H1 if the computed F-value is greater than 3.10. For the given sample data: The overall mean = x 13 15 18 15.792. 24 The sample means are: x.1 13 15 21 19 18.0 ; 6 12 15 16 18 x.3 13.5 ; x.4 17.333. 6 6 6 14.333 ; x.2 18.0 15.792 151.96. 2 2 SSTotal = (13.0 15.792) SST = 6 14.333 15.792 SSE = SS Total – SST = 151.96 - 87.79 = 64.17. 2 6 17.333 15.792 87.79. 2 Thus, we get the following ANOVA table: Source of Variation SS Factor Error Total df 87.79 3 64.17 20 151.96 23 MS F 29.26 9.12 3.21 The computed F-value is 9.12, which is greater than the critical value of 3.10. So, there is sufficient evidence, at = 0.05, to reject H0, that is, to infer that the mean number of crimes per day is not the same in all the districts. 19. Let 1, and 2 be the population means of the number of correct answers obtained by the two groups of students. Then, the null and the alternative hypotheses are: H0: 1 = 2 H1: 1 2. 12-9 (a) We have selected = 0.05. We shall use as test statistic F = MST . MSE Since the population distributions are approximately normal with equal variance, the distribution of the test statistic, under H0, is F-distribution with df = (k-1, n-k) = (1, 12). This is an upper-tailed F-test. For df = (1, 12), F = F0.05 = 4.75. The decision rule is: Reject H0 is favour of H1 if the computed F-value is greater than 4.75. For the given sample data: The overall mean is x 19 17 25 23.57. 14 The sample means are: x.1 19 16 6 19.00 ; x.2 SSTotal = (19.0 23.57) 2 32 25 8 27.00. 25 23.57 333.43. 2 SST = 6 19.0 23.57 8 27.0 23.57 219.43. 2 2 SSE = SS Total – SST = 333.43 – 219.43 = 114.00. Thus, we get the following ANOVA table: Source of Variation Treatment Error Total SS df 219.43 1 114.00 12 333.43 13 MS F 219.43 23.097 9.50 The computed F-value is 23.097, which is greater than the critical value of 4.75. So, there is sufficient evidence, at = 0.05, to reject H0, that is, to infer that the two means are different. b. We shall use two-tailed t-test. For df = (n1+n2-2)=12, t = t0.025 = 2.179. Hence, the decision rule is: reject H0 if t < 2 2.179 or if t > 2.179. t 19 27 1 1 9.5 6 8 4.806 . Since –4.806 < -2.179, there is sufficient evidence, at = 0.05, to reject H0, that is, to infer that the two means are different. c. 21. (4.806)2 23.097. So, t2 = F. Also (2.179)2 4.75. Hence, both the tests give us the same result: there is sufficient evidence, at = 0.05, to reject H0. For colour: 12-10 Let 1, 2, 3 and 4 be the population means of ratings given by the magazine readers for four different colours of advertisements. Then, the null and the alternative hypotheses are: H0: 1 = 2 = 3 = 4 H1: At least one of the ’s is different We shall use as test statistic: F = MST MSE The population distributions are approximately normal with equal variance. Hence, under H0, the test statistic has approximately an F-distribution with df = (k-1,n-k-r+1) = (3, 6). This is an upper-tailed F-test. We have selected a significance level of 0.05. For df = (3.6), F0.05 = 4.76. The decision rule is: reject H0 in favour of H1 if the computed F-value is greater than 4.76. For size: Let b1, b2, and b3 be the population means of ratings given by the magazine readers for three different sizes of advertisements. Then, the null and the alternative hypotheses are: H0: b1 = b2 = b3 H1: the means are not all equal. We shall use as test statistic: F = MST MSE The population distributions are approximately normal with equal variance. Hence, under H0, the test statistic has approximately an F-distribution with df = (r-1,n-k-r+1) = (2, 6). This is an upper-tailed F-test. We have selected a significance level of 0.05. For df = (2, 6), F0.05 = 5.14. The decision rule is: reject H0 in favour of H1 if the computed F-value is greater than 5.14. For the given sample data : The overall mean is x 23 8 5.5. 12 The sample means are: 23 6 35 7 3.67 ; x.2 5.0; 3 3 3 68 878 x.3 5.67 ; x.4 7.67. 3 3 x.1 x1. 2 8 4 4.0, x2. 3 SS(Color) 3 3.67 5.5 2 SS(Size) 4 4.0 5.5 2 SStotal 2 5.5 2 7 4 5.25, x3. 6 8 4 7.25. 3 7.67 5.5 25.0 2 4 7.25 5.5 21.5 2 8 5.5 55.0 2 SSE = SStotal - SS(Color) - SS(Size) = 55.0 – 25.0 – 21.5 = 8.5. Thus, we get the following ANOVA table: 12-11 Source of Variation SS df MS Color Size Error Total 3 2 6 11 8.33 5.88 10.75 7.59 1.42 25.0 21.5 8.5 55.0 F Our conclusions are as follows: Test for color: 5.88 is greater than the critical value, 4.76. Hence, there is sufficient evidence, at = 0.05, to reject H0, that is, to infer that the means for different colors are not all the same. Test for size: 7.59 is greater than the critical value, 5.14. Hence, there is sufficient evidence, at = 0.05, to reject H0, that is, to infer that the means for different sizes are not all the same. 23. Let 1, 2, 3 and 4 be the population means of the values of homes assessed by the four different assessors. Then, the null and the alternative hypotheses are: H0: 1 = 2 = 3 = 4 H1: the means are not all equal. We shall use as test statistic: F = MST MSE The population distributions are approximately normal with equal variance. Hence, under H0, the test statistic has approximately an F-distribution with df = (k-1,n-k-r+1) = (3, 12). This is an upper-tailed F-test. We have selected a significance level of 0.05. For df = (3, 12), F0.05 = 3.49. The decision rule is: reject H0 in favour of H1 if the computed F-value is greater than 3.49. For the given sample data : The overall mean = x 153 150 20 The sample means are: x.1 x.3 x1. x3. x5. 153 149 153 148 184 184 5 5 4 4 4 192 145 153 186 161.0 ; x.2 161.0 ; x.4 150.5; x2. 150.0; x4. 186 155 145 150 170 187.8. 12-12 161.30. 189 5 5 4 4 186 153 164 163.0; 160.2 151.5 ; 166.8 ; SST 5 161.0 161.3 5 160.2 161.3 21.4 SSB 4 150.5 161.3 4 187.8 161.3 4278.7 2 2 2 2 SS total 153.0 161.3 186.0 161.3 4428.2 2 2 SSE = 4428.2 – 21.4 – 4278.7 = 128.1. Thus, we get the following ANOVA table: SOURCE Home Assessor Error Total 4 3 12 19 DF 4278.7 21.4 128.1 4428.2 SS 1069.7 7.1 0.67 10.7 MS F 100.20 F = 0.67 is less than the critical value, 3.49. Hence, we do not have sufficient evidence, at = 0.05, to reject H0, that is, to infer that the population means are not all the same. 25. For Gasoline : Let 1, 2, and 3 be the population means of automobile fuel efficiencies using three different grades of gasoline – unleaded regular, mid-grade and super premium gasoline, respectively. Then, the null and the alternative hypotheses are: H0: 1 = 2 = 3 H1: The mean fuel efficiencies of the three grades of gasoline are not all the same. We shall use as test statistic: F = MST MSE The population distributions are approximately normal with equal variance. Hence, under H0, the test statistic has approximately an F-distribution with df = (k-1,n-k-r+1) = (2, 12). This is an upper-tailed F-test. We have selected a significance level of 0.05. For df = (2, 12), F0.05 = 3.89. The decision rule is: reject H0 in favour of H1 if the computed F-value is greater than 3.89. For Automobiles: Let b1, b2, b3, …, b be the population means of fuel efficiencies of the seven different automobiles. Then, the null and the alternative hypotheses are: H0: b1 b2 b7 H1: The mean fuel efficiencies of the seven automobiles are not all the same. For df = (6,12), F0.05 = 3.00. The decision rule is: reject H0 in favour of H1 if the computed F-value is greater than 3.00. For the given sample data: The overall mean = x 7.8 8.0 21 9.0 The sample means are: 12-13 8.2714. x.1 x.3 7.8 8.3 8.3 7 8.143 ; x.2 8.0 8.5 7 9.0 8.186; 8.486 ; 7 7.8 8.0 8.3 8.0 7.9 8.2 x1. 8.033 ; x2. 8.033 ; 3 3 8.1 8.2 8.5 8.1 8.1 8.3 x3. 8.267 ; x4. 8.167 ; 3 3 8.3 8.4 8.7 8.4 8.2 8.4 x5. 8.467 ; x6. 8.333 ; 4 3 8.3 8.5 9.0 x7. 8.6. 3 2 2 SS (Gasoline) 7 8.143 8.2714 7 8.486 8.2714 0.4886 SS ( Auto) 3 8.033 8.2714 2 SS total 7.8 8.2714 2 3 8.6 8.2714 0.8229 2 9.0 8.2714 1.4829 2 SSE = 1.4829-0.4886-0.8229=0.1714. We get the following ANOVA table: Autos Gasoline Error Total Source of Variation DF MS F 6 0.8229 0.1371 2 0.4886 0.2443 12 0.1714 0.0143 20 1.4829 SS 9.60 17.10 Our conclusions are as follows: Test for gasoline: 17.1 is greater than the critical value, 3.89. Hence, there is sufficient evidence, at = 0.05, to reject H0, that is, to infer that the mean fuel efficiencies, for the different grades of gasoline, are not all the same. Test for automobiles: 9.6 is greater than the critical value, 3.0. Hence, there is sufficient evidence, at = 0.05, to reject H0, that is, to infer that the mean fuel efficiencies of the automobiles are not all the same. 27. Let 1 and 2 be the population means of one-year percent returns on Canadian Balanced Funds and Health Care Companies, respectively. Then the null and the alternative hypotheses are: H0: 1 = 2 H1: 1 ≠ 2 12-14 Use as test statistic: F = MST MSE The population distributions are approximately normal with equal variance. Hence, under H0, the test statistic has approximately an F-distribution with df = (k-1,n-k-r+1) = (1, 4). This is an upper-tailed F-test. We have selected a significance level of 0.05. For df = (1, 4), F0.05 = 7.71. The decision rule is: reject H0 in favour of H1 if the computed F-value is greater than 7.71. Collect the random sample data, compute x , x.1 , x.2 , x1. , x2. , x3. , x4. , x5. , SSTotal, SST, SSB, and SSE, and construct the ANOVA table. If the computed F-value is greater than 7.71, conclude that there is sufficient evidence, at α = 0.05, to reject H0 in favour of H1. Else, conclude that we do not have sufficient evidence, at α = 0.05, to reject H0 in favour of H1. 29. (a) Let 12 and 22 be the population variances of the number of stolen bases among the teams that play their home games on natural grass and artificial turf, respectively. Then the null and the alternative hypotheses are: H0: 12 =1 22 H1: 12 1 22 This is a two-tailed test. Since the population distributions are approximately normal, we can use two-tailed F-test. The Minitab and Excel outputs are given below. MINITAB OUTPUT Test for Equal Variances Level1 C1 Level2 C2 F-Test (normal distribution) Test Statistic: 1.073 P-Value: 0.984 F-Test Two-Sample for Variances Variable 1 Variable 2 Mean 196.8696 188.2857 Variance 1911.937 1781.905 Observations 23 7 df 22 6 F 1.072974 P(F<=f) one-tail 0.508137 F Critical one-tail 3.8564 Both the outputs give us a P-value of 0.984. (This is directly given in Minitab output. From the Excel output we find P-value = min{2(0.508137), 2(1-0.508137)} = 0.983746 0.984. Since the value of α (=0.1) is smaller than the P-value, we do not have sufficient evidence to reject H0, that is, to infer that the two population variances are different. (b) Let 1, 2, and 3 be the population means of number of games won in the three groups. Then the null and the alternative hypotheses are: H0: 1 = 2 = 3 H1: The ’s are not all equal 12-15 The Minitab and Excel outputs are given below. From both the outputs, we get P-value = 0.039. Since the selected value of α (= 0.05) is greater than the P-value, we have sufficient evidence to reject H0, that is, to infer that the population means are not all the same. MINITAB OUTPUT Analysis of Variance Source DF SS Factor 2 618.5 Error 27 2273.4 Total 29 2891.9 MS 309.3 84.2 F 3.67 P 0.039 MICROSOFT EXCEL OUTPUT ANOVA Source of Variation SS Between Groups 618.502 Within Groups 2273.365 Total 2891.867 (c) df MS F P-value F crit 2 309.251 3.672872 0.038826 3.354131 27 84.19869 29 Let 1, 2, and 3 be the population means of team batting averages for the three groups. Then the null and the alternative hypotheses are: H0: 1 = 2 = 3 H1: The ’s are not all equal The Minitab and Excel outputs are given below. From both the outputs, we get P-value = 0.231. Since the selected value of α (= 0.1) is less than the P-value, we do not have sufficient evidence to reject H0, that is, to infer that the population means are not all the same. MINITAB OUTPUT Analysis of Variance Source DF SS Factor 2 0.000401 Error 27 0.003494 Total 29 0.003894 MS 0.000200 0.000129 F 1.55 P 0.231 MICROSOFT EXCEL OUTPUT ANOVA Source of Variation SS Between Groups 0.000401 Within Groups 0.003494 Total 0.003894 31. (a) df MS F P-value 2 0.0002 1.548684 0.230821 27 0.000129 29 F crit 2.51061 Let 12 and 22 be the population variances of the selling prices of the homes that do not have a pool and that have a pool, respectively. Then, the null and the alternative hypotheses are: 12-16 H0: 12 =1 22 H1: 12 1 22 This is a two-tailed test. Since the population distributions are approximately normal, we can use two-tailed F-test. The Minitab and Microsoft Excel outputs are given below. MICROSOFT EXCEL OUTPUT MINITAB OUTPUT Test for Equal Variances F-Test Two-Sample for Variances Level1 Level2 ConfLvl Mean Variance Observations df F P(F<=f) one-tail F Critical one-tail C1 C2 95.0000 F-Test (normal distribution) Test Statistic: 0.444 P-Value : 0.009 Variable 1 Variable 2 202.7974 231.4851 1136.031 2557.258 38 67 37 66 0.444238 0.004381 0.488537 Both the outputs give us a P-value of approximately 0.009. (From Excel output, we get the P-value for two-tailed test as 2(P(F<=f)) = 2(0.004381) = 0.008762 0.009.) Since the chosen value of α (=0.02) is greater than the P-value, we have sufficient evidence to reject H0, that is, to infer that the population variances are not equal. (b) Let 12 and 22 be the population variances of the selling prices of the homes that do not have and that have an attached garage, respectively. Then, the null and the alternative hypotheses are: H0 : 12 1 22 H1 : 12 1 22 This is a two-tailed test. Since the population distributions are approximately normal, we can use two-tailed F-test. The Minitab and Microsoft Excel outputs are given below. MINITAB OUTPUT Test for Equal Variances Level1 Level2 C1 C2 F-Test (normal distribution) Test Statistic: 0.389 P-Value : 0.004 MICROSOFT EXCEL OUTPUT F-Test Two-Sample for Variances Variable 1 Variable 2 Mean 185.45 238.1761 Variance 784.2571 2013.896 Observations 34 71 df 33 70 F 0.389423 P(F<=f) one-tail 0.001871 12-17 F Critical one-tail 0.47369 Both the outputs give us a P-value of approximately 0.004. Since the chosen value of α (=0.02) is greater than the P-value, we have sufficient evidence to reject H0, that is, to infer that the population variances are not equal. (c) Let 1, 2, and 3 be the population means of selling prices among the five townships. Then, the null and the alternative hypotheses are: H0: 1 = 2 = 3 H1: the population means are not all equal. Since the population distributions are approximately normal with equal variances, we shall use ANOVA test. The Minitab and Microsoft Excel outputs are given below. MINITAB OUTPUT Analysis of Variance Source DF SS Factor 4 13263 Error 100 217505 Total 104 230768 MS 3316 2175 F 1.52 P 0.201 MICROSOFT EXCEL ANOVA Source of Variation SS Between Groups 13262.85 Within Groups 217504.7 Total 230767.6 df MS F P-value F crit 4 3315.711 1.524432 0.200824 3.061984 100 2175.047 104 Both the outputs give a P-value approximately equal to 0.201. Since the selected value of (= 0.02) is less than the P-value, we do not have sufficient evidence to conclude that the population means are not all equal. 12-18