11.5 Tests about a Population Mean: Small Sample Case Example: Objective: determine whether Heathrow airport provides superior service as n 12, x 7.75, s 1.215 and 0.05 . That is, test the hypothesis H 0 : 0 7 vs. H a : 0 7 . Intuitively, if H 0 : 7 is true, for example, as the sample mean x 7 , it is very likely that should have value close to 7. Thus, a sensible statistical procedure would be where c reject H 0 : x 7c not reject H 0 : x 7c is some constant. Note that the above test is equivalent to reject H 0 : not reject H 0 : where c c s x 7 c s n x 7 c s n c 1.215 n is some constant. 12 Similar to previous section, we can use level of significance to obtain the constant c . As H 0 is true, for ease of explanation, suppose 7, X 0 X 7 ~ T n 1 T 11 . S S n 12 Then, 1 0.05 the probabilit y that wro ngly reject H 0 P( H 0 is true but is rejected) X 7 P c P T (11) c S 12 t c tn1, t11, 0.05 and c 1.215 . 12 11,0.05 Thus, a sensible statistical procedure would be x 7 reject H 0 : t 1.215 12 not reject H 0 : t t n1, x 0 t n1, t11, 0.05 s n with X 7 P 7; reject H 0 P t11, 0.05 ; 7 0.05 S n Thus, t x 7 1.215 12 7.75 7 1.215 12 2.14 t11,005 1.796 . Therefore, we reject H 0 . The other approach is to use p-value. p-value is the probability of making a type I error as reject H 0 at x , i.e., X 0 x 0 p - value P , 7 PT (n 1) t PT (11) 2.14 0.028. s S n n Since p value 0.05 , we reject H 0 . 2 General Case: as n 30 and level of significance t H 0 : 0 vs. x 0 x 0 sX s . n H a : 0 Then, reject H 0 : t t n 1, not reject H 0 : t t n 1, In addition, p - value PT (n 1) t H 0 : 0 vs. H a : 0 Then, reject H 0 : t t n 1, not reject H 0 : t t n 1, In addition, p - value PT (n 1) t . H 0 : 0 vs. H a : 0 Then, reject H 0 : t t n 1, not reject H 0 : t t n 1, 3 2 2 In addition, p - value PT (n 1) t Example 1: Objective: determine if the mean filling weight is exactly 16 ounces. That is to test H 0 : 0 16 vs. H a : 0 16 with the following data 16.02 16.22 15.82 15.92 16.22 16.32 16.12 15.92 Suppose the level of significance is 0.05. [solution:] n 8, x 16.07, s 0.18 . Thus, t x 0 16.07 16 1.10 . s 0.18 n 8 Since t 1.10 1.10 t n 1, t 7,0.025 2.365 , 2 we do not reject H0 . Example 2: A sample of 6 observations, 18, 20, 16, 19, 17, 18, is taken. Suppose the population is normally distributed. Then, (a) for the hypothesis, H 0 : 20 vs. H a : 20 , using 0.05 , test the hypothesis based on the classical hypothesis test? (b) for the hypothesis, H 0 : 18 vs. H a : 18 using 0.05 , test the hypothesis based on p-value? (c) for the hypothesis, 4 H 0 : 17 vs. H a : 17 using 0.05 , test the hypothesis based on the confidence interval approach? [solutions:] 6 n 6, x 18 20 16 19 17 18 18, s 6 x x i 1 2 i 6 1 2. (a) 0 20, t x 0 18 20 3.464 t n1, t 5, 0.05 2.015 reject H 0 s 2 n 6 (b) 0 18, x 0 18 18 0 p - value PT (n 1) t PT (5) 0 0.5 0.05 s 2 n 6 not reject H 0 t (c) A 95% confidence interval for is x t n1, 2 s 2 18 t 5,0.025 18 2.571 0.577 16.516,19.483 n 6 Since 17 16.516,19.483 , we do not reject H 0 . Example 3: Consider the following hypothesis test, H 0 : 50 vs. H a : 50. . Assume a sample of 16 items provides the following test statistics. Use 0.05 , test the hypothesis based on p-value as (i) t 1.055 (ii) t 3.261 . [solution:] (i) n 16, p - value PT (n 1) t PT (15) 1.055 PT (15) 1.753 0.05 do not reject H 0 5 (ii) n 16, p - value PT (n 1) t PT (15) 3.261 PT (15) 1.753 0.05 reject H 0 Online Exercise: Exercise 11.5.1 Exercise 11.5.2 6