Survey
* Your assessment is very important for improving the workof artificial intelligence, which forms the content of this project
* Your assessment is very important for improving the workof artificial intelligence, which forms the content of this project
IME 301 Goodness-of-fit Testing Examples By: Dr. Parisay This handout presents discussion on Goodness-of-fit testing examples. The detail solutions were performed in class. Dr. Parisay’s comments are in red Problem : The grades in a statistics course for a particular semester were as follows: Grade F A 14 B 18 C 32 D 20 F 16 Test the hypothesis at the 0.5 level of significance, that the distribution of grades is uniform. Solution: H0 = data has uniform distribution H1 = data has not uniform distribution n = number of observations = 14+18+32+20+16 = 100 k = number of intervals = 5 for a uniform distribution ei = expected number of observations in each interval = n/k = 100/5 = 20 p = number of parameters of fitted distribution (uniform distribution parameters: min and max) = 2 2,( k p1) 02.05,2 5.991 02 (14-20)2/20 + (18-20)2/20 +(32-20)2/20 +(20-20)2/20 +(16-20)2/20 = 200/20 = 10 02 10 > 5.991 conclude cdf = uniform is not a good fit Version 2: The grades in a statistics course for a particular semester were as follows: Grade A B C D F F 14 18 32 20 16 Test the hypothesis at the 0.5 level of significance, that the distribution of grades follows a normal distribution with 15% for A, 20% for B, 30% for C, 20% for D, and 15% for F. In fact 15% for A means that the area under the normal curve for the range of grades (above 90%) that will receive A is 0.15, or probability of receiving an A is 15%. Using probability notation [F(max grade for A: 100%) – F(min grade for A: 90%)], where F(X) is cumulative distribution function for variable X. 1 Solution: n = number of observations = 14+18+32+20+16 = 100 Interval Oi ei = probability * n A B C D F 14 18 32 20 16 15%(100) =15 20%(100) =20 30%(100) =30 20%(100) =20 15%(100) =15 Total (Oi - ei )2 / ei (14-15)2 /15 (18-20)2 /20 (32-30)2 /30 (20-20)2 /20 (16-15)2 /15 0.47 p = number of parameters of fitted distribution (normal distribution parameters: mean and std. dev.) = 2 k = number of intervals = 5 2,( k p1) 02.05,2 5.991 02 0.47 < 5.991 conclude cdf = normal under the above condition is a good fit Version 3: The grades in a statistics course for a particular semester were as follows: Grade A B C D F F 14 18 32 20 16 Test the hypothesis at the 0.5 level of significance, that the distribution of grades follows a normal distribution with 10% for A, 20% for B, 40% for C, 20% for D, and 10% for F. Solution: n = number of observations = 14+18+32+20+16 = 100 Interval Oi ei = probability * n A B C D F 14 18 32 20 16 10%(100) =10 20%(100) =20 40%(100) =40 20%(100) =20 10%(100) =10 Total (Oi - ei )2 / ei (14-10)2 /10 (18-20)2 /20 (32-40)2 /40 (20-20)2 /20 (16-10)2 /10 7 p = number of parameters of fitted distribution (normal distribution parameters: mean and std. dev.) = 2 k = number of intervals = 5 2,( k p1) 02.05,2 5.991 02 7 > 5.991 conclude cdf = normal under the above condition is not a good fit Date prepared: November 29, 2003 2