2/27/98 252y9812 ECO252 QBA2 FIRST HOUR EXAM February 19, 1998 Name Hour of Class Registered (Circle) MWF 10 11 TR 12:30 2:00 Hour of Class Attended (If Different) ______________ I. (14 points) Do all the following. x ~ N 2,6.5 1. 2. 30 2 1 2 z . z 4.31 P1 x 30 P P 015 6.5 6.5 P 015 . z 0 P 0 z 4.31 .0596.5000 .5596 3 2 3 2 z . P 3 x 3 P P 0.77 z 015 6.5 6.5 P 0.77 z 0 P 0 z 015 . .2794.0596 .3390 3. 6 2 22 z P 2 x 6 P P 0 z 0.62 .2324 6.5 6.5 4. 3 2 P x 30 . P z P z 0.77 6.5 P 0.77 z 0 P z 0 .2794.5000 .7794 0 2 3 2 z P 30 . x 0 P P 0.77 z 0.31 6.5 6.5 P 0.77 z 0 P 0.31 z 0 .2794.1217 .1577 5. 6. A symmetrical interval about the mean with 93% probability. We want two points x .0350 and x.9650 , so that P x.9650 x x .0350 .9300 . From the diagram, if we replace x by z, P 0 z z.0350 .4650 . The closest we can come is . , and x z.0350 2 181 P 0 z 181 . .4649 . So z.0350 181 . 65 . 2 1177 . , or -9.77 to 13.77. 7. x .07 We want a point x .07 , so that P x x .07 .07 . From the diagram, if we replace x by z, P 0 z z.07 .43 . The closest we can come is P0 z 148 . .4306 or , and . P0 z 147 . .4292 . Use something between the two. So z.07 1475 x z.07 2 1475 . 65 . 2 9.59 , or 11.59. 1 2/25/98 252y9812 II. (6 points-2 point penalty for not trying part a.) Show your work! A sample of seven printheads were tested for durability. The number of characters (in millions) printed before failure are listed below. Assume that we were sampling from a normal distribution. Printer Characters 1 0.8 2 1.4 3 1.3 4 1.1 5 0.9 6 1.7 7 1.3 a. Compute the sample variance of the above numbers. Show your work.(3) Solution: Printer x x2 x 8.5 (Characters) x 12143 . 1 0.8 0.64 n 7 2 1.4 1.96 2 2 2 x nx 10.89 712143 . 2 3 1.3 1.69 s n 1 6 4 1.1 1.21 Unfortunately, with 009476 . 5 0.9 0.81 a bit too much rounding, you can get 6 1.7 2.89 7 1.3 1.69 . . s 2 .1967 , which gives s x 01235 Total 8.5 10.89 b. Compute a 98% confidence interval for the mean time.(3) From the problem statement .02 . From page 10 of the syllabus supplement, if the . population variance is unknown x t s x and t n21 t .601 3143 . Solution: 2 sx s n 0.09436 0.3078 . So 12143 0116350 . . 3143 . 0116350 . . 0.37 or 0.85 to 1.58. 121 7 7 2 2/25/98 252y9812 III. Do at least 2 of the following 4 Problems (at least 10 each) (or do sections adding to at least 20 points Anything extra you do helps, and grades wrap around) . Show your work! 1. Using the data from the previous problem (page 2.), test the hypothesis that the population mean is 1.0 at the 90% confidence level using: a. A test ratio (2) b. Critical values (2) c. A confidence interval (2) d. Find an approximate p-value for the null hypothesis. (1) e. Now find a 90% confidence interval for the standard deviation.(3) Solution: From page 10 of the Syllabus Supplement: Interval for Confidence Interval Hypotheses Mean ( known) x z x Mean ( unknown) x t sx 2 2 DF n 1 H 0: 1 H 1: 1 Test Ratio H0 : 0 H1 : 0 H 0 : 0 H1: 0 t x 0 x x 0 x cv 0 z 2 x x cv 0 t 2 s x sx n 1 6 0 1 DF n 1 6 .10 t t .05 1943 . 2 From the previous page: x 12143 and s x . a. Test Ratio: t z Critical Value s n 0.09436 0.3078 . 0116350 . 7 7 x 0 12143 . 1 = 1.8419. This is in the 'accept region' sx .116350 between -1.943 and +1.943, so accept H 0 . b. Critical Value: x cv 0 t s x 1 1943 . 0116350 . 1 0.226 or 0.774 2 to 1.226. This includes x 12143 , so accept H 0 . . c. Confidence Interval: x t sx 12143 . 1943 . 011635 . . 0.226 1214 2 or 0.988 to 1.440. This includes 0 1 , so accept H 0 . 6 6 . . d. From the t table t = 1.842 is between t .05 1943 and t .10 1440 . Since this is 2-sided, we double pvalue, and .10 p value .20 . e. From page 1 of the Syllabus Supplement: n 1 s 2 2 2 2 2 6 2 6 Since .05 and 1 .95 , look up .05 and .95 . So 2 2 6 0.09476 2 12.5916 0.2124 05496 . 6 0.09476 16354 . n 1 s 2 2 1 . DF n 1 6 and .10 . 2 n 1 s 2 .205 2 n 1 s 2 .295 or or 0.04515 2 0.3427 . Finally, taking square roots, 3 2/25/98 252y9812 2. Data from section II is repeated below. Printer Characters 1 0.8 2 1.4 3 1.3 4 1.1 5 0.9 6 1.7 7 1.3 a. Test that the population mean is 1 assuming that the parent population is normally distributed and that the sample of seven is taken from a population of 50. (3) b. Drop the assumption that the parent distribution is normal ( and that the populate size is 50) and figure out a confidence level for the confidence interval ( 08 . 17 . ) (3) c. Put a p-value on the statement that the median is 1, again assuming that the distribution is not normal. (2) d. Looking at your result in c, would you reject the hypothesis that the median is 1 at the 90% confidence level? Explain. (1) e. Repeat c-d assuming that the parent distribution is normal.(2) Solution: n 7, N 50 , and from Part II s 2 0.09476 and x 12143 . . a. s x s n N n N 1 s2 N n n N 1 0.09476 50 7 . 0.011880 010899 50 1 7 x 0 12143 . 1 = 1.966. This is not in the 'accept region' sx .10899 (i) Test Ratio: t between -1.943 and +1.943, so reject H 0 . or (ii) Critical Value: x cv 0 t s x 1 1943 . 010899 . 1 0.212 or 0.788 2 to 1.212. This does not include x 12571 , so reject H 0 . . or (iii) Confidence Interval: x t sx 12571 . 1943 . 010899 . . 0.212 1257 2 or 1.045 to 1.469. This does not include 0 1 , so reject H 0 . b. Since 0.8 and 1.7 are the highest and lowest numbers, the interval is wrong if all seven numbers are below the median or all seven are above the median. The first is equivalent to getting seven heads in seven tosses of a fair coin, and the second is equivalent to getting seven tails but both these probabilities are the same . So, using the binomial table, 2 P x 7 n 7, p .5 21 P x 6 21.99219 2.0078 .01562 and the confidence level is 1 .9843 . H0 : 1 c. There are two numbers below 1 and five numbers above 1. A p-value for the statement that H1 : 1 the median is 1 would be the probability of 5 or more numbers above the median, equivalent to the probability of getting 5 or more heads.. This is doubled because the statement is 2-sided. p value 2 P x 5 n 7, p .5 2 1 P x 4 21.63672 2.3633 .7266 d. Since the p-value (.7266) is greater than the significance level (.10) accept H 0 . e. Since , if the distribution is normal, the mean and the median are equal, this is the same as a on this or the previous page. 4 2/25/98 252y9812 3. A manufacturer of condensers says that fewer than 4% of the condensers produced are defective. A sample of 300 is selected from a very large shipment and 10 defective items are found. a. Do a 2-sided 95% confidence interval for the proportion that are defective.(2) b. Do a 2-sided 93% confidence interval for the proportion that are defective.(2) c. State the Hypotheses to be tested in the problem above.(1) d. Do a test of the hypothesis in c) at a 95% confidence level. (3) e. Repeat d) at the 93% confidence level.(2) f. If the true proportion defective is 3%, what is the power of the test? (3) g. The New York Times/CBS News poll says that 10% of the population believe the President Clinton is telling the whole truth about Paula Jones. On how large a sample would this have to be based to be accurate within 0.5%? (2) Solution: From page 10 of the Syllabus Supplement: Interval for Confidence Interval Hypotheses Test Ratio Proportion p p z 2 s p H 0 : p p0 H 1 : p p0 z pq n q 1 p sp p p0 p Critical Value p cv p 0 z 2 p p p0 q 0 n .0333.9667 pq pq x 10 q 1 p 1.0333 .9667 s p .0333 s p .010364 . n n 300 n 300 . .010364 .0333 .0203 or .0130 to .0536. a. .05 .025 p p z.025 s p .0333 1960 2 2 You do not need a null hypothesis for a confidence interval unless you are specifically using it to test the hypothesis! .07 .035 p p z.035 s p .0333 181 . .010364 .0333 .0187 or .015 to .052. This value b. 2 2 of z was found on page 1. H : p .04 c. 0 H 1 : p .04 p d. .05 p0 .04 q 0 1 p0 .96 p p0 q 0 n .04.96 300 .0113137 . This is a one-sided test with .05, z z.05 1645 . . p p 0 .0333.04 0.5893 . Since this is above (i) Test Ratio: z p .0113137 -1.645, accept H 0 . . .0113137 .02138 . or (ii) Critical Value: pcv p0 z p .04 1645 Since p .0333 is above this, accept H 0 . . .013137 or or (iii) Confidence interval: p p z s p .0333 1645 p .05194 . This does not contradict H 0 p .04 , so accept H 0 . 5 2/24/98 252y9812 e. This is a one-sided test with .09, z z.09 1475 . This value of z was found on page 1. . p p 0 .0333.04 0.5893 . Since this is above (i) Test Ratio: z p .0113137 -1.475, accept H 0 . . .0113137 .0233 . or (ii) Critical Value: pcv p0 z p .04 1475 Since p .0333 is above this, accept H 0 . . .013137 or or (iii) Confidence interval: p p z s p .0333 1475 p .0500 . This does not contradict H 0 p .04 , so accept H 0 . f. Use a critical value from d or e. In d, we accept H 0 if p is greater than . .0113137 .02138 . We will compute , the probability of a type II error . pcv p0 z p .04 1645 p p1 .03.97 p1q1 by finding P p p cv p p1 P z cv where p1 .03 and p1 p1 n 300 .02138.03 .0.00985. P p .02138 p .03 P z P z 088 . P 088 . z 0 P z 0 .00985 .3106.5000 .8106 . g. If .05 , z z.025 1960 and n . 2 pqz 2 e2 2 . .10.90196 .005 2 13829.8 , so use a sample of size 13830. 6 2/25/98 252y9812 4. A sample variance is found to be 250. The parent distribution is approximately normal. a. Do a 98% confidence interval for the population variance if the sample size is 28. (2) b. Do a 98% confidence interval for the population variance if the sample size is 125. (2) c. If the sample size is 125, do a test of the statement that the population variance is 200 at the 98% confidence level using the method taught in class. (3) d. Repeat the test in c) using critical values for the sample variance as explained on page 10 of the syllabus supplement. (2) e. Looking at what you did in d), do a test of the statement that the population variance is at least 200. (3) f. If we claim that a population has a Poisson distribution with a mean of 7 , and the actual value found is 4, do a hypothesis test at the 95% level. (3) Solution: s 2 250 and from page 10 of the Syllabus Supplement: Interval for Confidence Hypotheses Test Ratio Critical Value Interval VarianceSmall Sample 2 VarianceLarge Sample n 1s 2 H 0 : 2 02 .25.5 2 H1: : 2 02 s 2DF H 0 : 02 H1 : 2 02 2 2 scv 02 .25.5 20 2 n 1 z 2 z 2 2DF n 1s 2 2 2 2DF 1 a. The variance formula is explained on pages 1 and 2 of the Syllabus Supplement. .02 and there are 27 degrees of freedom. 27 250 n 1s 2 2 2 n 1s 2 2 1 2 27 250 becomes n 1 s 2 .201 2 n 1 s 2 .299 and then 2 or 143.73 2 524.31 . 2 46.9631 12.8785 b. .02 and there are 124 degrees of freedom. Since this is too many degrees of freedom for the 2 table, use the large sample formula. s 2 DF z 2 DF 2 250 2124 2.326 2124 250 2124 2.326 2124 s 2 DF z 2 DF where z z.01 2.326 . 2 2 . 18551 . . If we square this, it becomes or 13776 . 189.794 2 344157 . 2 H 0 : 200 2 n 1 s 2 124 250 155 . This is a large sample. z 2 2 2 DF 1 c. 2 2 200 0 H 1 : 200 2155 2124 1 17.6068 157162 . 189 . . .01 , z z.01 2.326 2 and z is between 2.326 , so accept H 0 . 7 2/24/98 252y9812 .25.5 20 H 0 : 2 200 2 2 2 d. so 0 200 . As the table (above) says, scv . The two values of 2 n 1 H 1 : 200 .25.5 are 12 .299 12.8785 and 2 .201 46.9631 .One of these values is 2 2 2 2 scvL 12.8785 200 2 95.40. , and the other is scvU 27 between them, accept H 0 . 46.9631 200 27 347.81 . Since s 2 250 is H 0 : 2 200 e. H 1 : 2 200 (i) Critical Value: Use 2 scv 2 with .01 , so scv .25.5 20 2 n 1 12 20 , but this is a one-sided test .298 20 14.01 250 130 . n 1 n 1 27 There was trouble here, but no one called it to my attention. The 2 2 table does not give .98 , but does give .975 , which is about 14.5, and 2 would be acceptable. I looked at .99 and settled on 14.01, which is about two-thirds of the way between them. Since s 2 250 is above 130, accept H 0 . or(ii) Test Ratio: 2 n 1 s 2 2 0 27 250 200 33.7500 . Since this is above .298 14.01 , accept H 0 . H 0 : Poisson 7 f. This is the easiest problem on the exam. This is a 2-sided test of . Actually, we are H 1 : not Poisson 7 assuming Poisson and checking for a mean of 7. The p-value of the null hypothesis when x 4 (4 is below the mean) is (from the Poisson table with a parameter of 7) 2 P x 4 2 .17299 = .3460. Since this is above the significance level (.05), accept H 0 . 8 Note: Last year we used a take-home exam, which we probably will never do again. This is one of the 10 versions of the exam. A shorter version of these problems is fair game. 252y981b 2/20/98 Name __________________ Hour of Class Registered (Circle) MWF 10 11 TR 12:30 2 Hour of Class Attended (If Different) __________ First Take-home Exam - QBA2 PART IV. Show your work! Neatness counts (Especially in Graphs) 1. The Trash Bag Case: (Bowerman and O'Connell) A television network refuses to air a commercial that claims that a trash bag has a breaking strength of 50 lbs. without proof. A sample of 250 shows that x 12644 and x 2 640156 . Do a 2-sided test of the claim using a confidence level of 99%. a. State the H 0 and H1 (1) b. Using critical values for the sample mean, test the Hypotheses. (2) c. Construct a power curve for the test. (7) Solution: x 12644 x 2 nx 2 640156 25050.5760 2 2 . x 50.5760 . s 2.7030 . s 16441 . n 250 n 1 249 H 0 : 50 a. n 250, .01 H 1 : 50 b. From page 10 of Syllabus supplement x cv 0 t s x . 0 50 . Since this is a large sample 2 t 2 z 2 z.025 2.576 . s x s 2.7030 . So xcv 0 t s x .010812 0103981 . 2 250 n is not between 50 2.576 0103981 . . 50 0.2679 or xcvl 49.732 to x cvu 50.268. Since x 505760 these values, reject H 0 . c. We accept H 0 if x is between 49.732 and 50.268. We will compute , the probability of a type II error x 1 x 1 by finding P x cvl x x cvu 1 P cvl z cvu for the values 1 0 , x x 0 0.5 t 2 s x , 0 10 . t 2 s x , 0 15 . t 2 s x and 0 2.0 t 2 s x , where t 2 s x 2.576 0103981 . 0.2679 . The values for this appear below. All computations are for 1 upper . 1 lower 1 upper zL x cvl 1 x zU 50.00 50.00 -2.58 49.87 50.13 -3.86 49.73 50.27 -5.15 49.60 50.40 -6.44 49.46 50.54 -7.72 * Pz L z zU with values from the z or t table. x cvu 1 x 2.58 1.29 0.00 -1.29 -2.58 * power 1 .9900 .9014 .5000 .0985 .0050 .0100 .0980 .5000 .9015 .9950 9 252y981b 2/20/98 2. The Cheese Spread Case: (Bowerman and O'Connell) (Extra credit) A cheese spread producer will not adopt a new package unless there is strong evidence that fewer than 10% of users will stop buying the product if the new package is used. A sample of 1000 is taken and 63 consumers say that they will not buy the product. State the hypotheses, do the test at the 95% confidence level and construct the power curve for the test. (5) You can't answer a problem you haven't read. If you can't answer the questions below, you didn't read the problem! 1. Where in this problem is a mean mentioned? 2. Which of your hypotheses is 'fewer than 10%' (not 'at most 10%')? Solution: H : p .10 a) 0 H 1 : p .10 p0 .10 n 1000 x 63 .063 .05 n 1000 Common error: the statement says 'fewer than.' This must be in your hypotheses and since it does not contain an equality, it cannot be a null hypothesis. p0 q 0 b) From page 10 of Syllabus supplement pcv p0 z 2 p where p . This must become n and . pcv p0 z p because the hypothesis is one sided. z.05 1645 p .10.90 p0 q 0 . .00949 .0844 . We will accept 0.00949 . So pcv p0 z p .10 1645 n 1000 H 0 if p is greater than .0844. Since p .063 , reject H 0 . c) We accept H 0 if p is greater than .0844. We will compute , the probability of a type II error by p p p1 finding P p p cv p p1 P z cv where p1 p1 p1q1 for the values p1 p0 , n . z 2 p and p0 2.0 z 2 p , where p0 0.5 z 2 p , p0 10 . z 2 p , p0 15 z 2 p 1645 . .00949 .0156 . The values for this appear below. p1 p1 .1000 .0922 .0844 .0766 .0688 .00949 .00915 .00879 .00841 .00800 * P z z1 z1 * power 1 -1.645 .9500 .0500 -0.853 .8023 .1977 0.000 .5000 .5000 0.928 .1762 .8238 1.950 .0256 .9744 p p1 where z1 cv computed from rounded z 1 using the z or t table. p1 10