252solngr3-061 11/01/06 (Open this document in 'Page Layout' view!) Name: Class days and time: Please include this on what you hand in! Graded Assignment 3 In your outline there are 6 methods to compare means or medians, methods D1, D2, D3, D4, D5a and D5b. Methods D6a and D6b compare proportions and method D7 compares variances or standard deviations. In the following cases, identify H 0 and H 1 and identify which method to use. Method E1 is a chi-squared test that compares multiple proportions. Method F1 is analysis of variance (ANOVA), which compares multiple means. If the hypotheses involve two means, state the hypotheses in terms of both and 1 2 . If the hypotheses involve two proportions, state them in terms of both p and p p1 p 2 . If the hypotheses involve two standard deviations or variances, state them in terms of both 2 and 12 22 or 22 12 . All the questions involve means, medians, proportions or variances. One of these problems is a chi-squared test. Note: Look at 252thngs (252thngs) on the syllabus supplement part of the website before you start (and before you take exams). Neatness and clarity of explanation are expected. Note that from now on neatness means paper neatly trimmed on the left side if it has been torn, multiple pages stapled and paper written on only one side. ----------------------------------------------------------------------------------------------------------------------------Example: This may seem long but it appears on a previous graded assignment 3. A group of supervisors are given exams on management skills before and after taking a course in management. Scores are as follows. Supervisor Before After 1 63 78 2 93 92 3 84 91 4 72 80 5 65 69 6 72 85 7 91 99 8 84 82 9 71 81 10 80 87 11 68 93 If we assume that the distribution of results is Normal, what method should we use to answer the question “Has the course improved the scores of the managers?” What are our hypotheses? Solution: You are comparing means before and after the course. You can get away with using means because the parent distributions are Normal. If 2 is the mean of the second sample, you are hoping that 2 1 , which, because it contains no equality is an alternate hypothesis. So your hypotheses are H 0 : 1 2 H 0 : 1 2 0 H 0 : D 0 or . If D 1 2 , then . The important thing to notice H : H : 0 2 2 1 1 1 1 H 1 : D 0 here is that the data are in before and after pairs, so you use Method D4. 252solngr3-061 11/01/06 1. West Chester University plays 11 football games a year. Though it is a questionable practice, these 11 games can be regarded as a random sample taken from an infinite number of games that they might have played. Statistics can be computed from them like proportion of games won or the mean and standard deviation of touchdowns or completed passes per game. At Christmas 2007, you look at these statistics and assert that the team did significantly better in 2007 than it did in 2006. a. The statistic you wish to use is completed passes. If x1 is a column listing the number of passes completed in each game in 2006 in random order and x 2 is a column listing the number of passes completed in each game in 2007 in random order and you consider these as independent random samples and wish to compare the mean passes completed per game and to get results that verify your statement, what are your hypotheses? What is your method? b. Your roommate, Gigglebutz, says that you have done this all wrong, because the distributions have not been shown to be Normal. You do the test described in section 6.3 of the text and sadly agree with Gigglebutz that the distributions are probably not Normal. What is your test and method now? c. Gigglebutz is still not satisfied and points out that these are not independent random samples and what you should do is rearrange all your data so that each column represents a year and each row an opponent. Any school that was not played in both years should be dropped from your data. What is your test and method now? 2. Standard deviation is often a measure of reliability. A manufacturer is providing a connector and is getting complaints that the connector is often too large or too small for the intended use. A sample of 250 connectors produced by the process in current use is taken. Then a new process is tried and a sample of 45 connectors is taken. Find the hypotheses and method to show that the new process produces a more reliable product. 3. (Dummeldinger pg 148) We are interested in attitudes about sexual discrimination. Various groups were asked if they believed sexual discrimination is a problem in the United States. .05 a. Out of 50 men 11 believed that sexual discrimination is a problem. Out of 40 women 19 said that sexual discrimination is a problem. Is there a significant difference between male and female attitudes? b. A group of 100 men are picked from upper levels of a group of large American firms. They are asked if sexual discrimination is a problem in the United States. 38 said ‘yes.’ They are then sent to a program on mentoring female executives where they discuss, among other things, what problems seem to be peculiar to young female executives. Afterwards the men are asked the same question. This time 41 said ‘yes.’ Assuming that these 100 men are representative of all executives who are trained to mentor female executives, has the training increased the proportion that believes there is a problem? c. Random samples (each 100 people) of male 1) high school graduates, 2) college graduates and 3) people with MBAs are asked the same question. Out of the first group 35 said ‘yes.’ Out of the second group 38 said ‘yes.’ Out of the third group 39 said ‘yes.’ Is there a significant difference between the attitudes of the three groups? Extra credit: (Place answers at the end of the assignment) d. You have enough information to do two of the three problems above. What additional information do you need? e. Do one of the three problems. State your conclusion clearly. 4. (Dummeldinger) A researcher took a random sample of n graduates of MBA programs, which included n1 women and n 2 men. Their starting salaries were recorded. Use 1 and/or 1 for population parameters for women and 2 and/or 2 for women. Choose hypotheses and methods for the following. .10 a. n1 n 2 150 . The researcher wants to show that men have a higher mean starting salary than women. b. n1 18 and n2 12 . The researcher wants to compare the dispersion of men’s and women’s salaries. The researcher has no prior opinion as to which is more variable. 2 252solngr3-061 11/01/06 c. n1 18 and n2 12 and the null hypothesis in 4b) has not been rejected. Dummeldinger gives the following data. x1 48266 .70 x 2 55000 .0 and . The researcher wants to show that men have a s1 13577 .63 s 2 11741 .25 higher mean starting salary than women. Extra credit: (Place answers at the end of the assignment) d. Do problem 4c. (Hint: Before you start, move the decimal point so that the sample means and standard deviations are in thousands.) State your conclusion clearly. 5. (Dummeldinger) In order to find the effect of membership in a fraternity or sorority on grades, 250 students who had joined fraternities or sororities were picked as a random sample. a. GPA’s were recorded for each student for the semester before ( x1 column) and the semester after they joined ( x 2 column). Mean GPAs were compared. Choose hypotheses and method. b. Dummeldinger says that a 95% confidence interval for the difference between the means was 0.24 to 1.02. What is your conclusion?) Solutions General considerations. 1) All methods in section D are methods that can only be used for comparison of 2 samples. This is because, if (theta) is a parameter like or p, 1 2 is easy to define and will be zero if 1 and 1 are equal. If we go to more than two samples, say 3, we need something like 1 0 2 2 0 2 3 0 2 , where 0 is some sort of average of the parameters of the samples. This will equal zero if all the parameters are equal and will not allow positive discrepancies in one sample to cancel out negative discrepancies in another. This is what takes us to chi-squared and ANOVA methods. Saying 2 2 2 0 is not the same as saying D 3 0 , because 1 0 2 0 3 0 1 2 3 D 3 would be negative if 1 2 3 , but saying 1 0 2 2 0 2 0 is the same as saying D 2 0 . (Try proving this – it’s simple algebra.) 1 2 2) You can always substitute a method for the median for a method for the mean, but not vice versa. However, if a Normal distribution applies, a method involving means will be more efficient and powerful. 3) The computer will used Method D3 when it is not told what method to use. This is quite general because if the sample variances are similar, it gives results like D2 and if the sample sizes are large, it gives results like D1. However, if variances are equal D2 is easier to use and if the samples are large D1 is easier to use. 4) The K-S and Lilliefors methods only exist because chi-square performs so poorly for small samples. K-S needs , or other parameters. Lilliefors uses x or s and only works to test for a Normal distribution. 5) ‘Significant’ in statistics means that we have rejected a hypothesis like H 0 : 0 and ‘significantly different’ means that we have rejected a hypothesis like H 0 : 1 2 . Of course, if two parameters are significantly different, their difference is significant. 6) Be careful of inequalities. If 1 2 or 2 1 and D 1 2 , then D 0. 7) In most problems you are better off trying to figure out what the alternative hypothesis is before you try to state the null hypothesis. 3 252solngr3-061 11/01/06 1. West Chester University plays 11 football games a year. Though it is a questionable practice, these 11 games can be regarded as a random sample taken from an infinite number of games that they might have played. Statistics can be computed from them like proportion of games won or the mean and standard deviation of touchdowns or completed passes per game. At Christmas 2007, you look at these statistics and assert that the team did significantly better in 2007 than it did in 2006. a. The statistic you wish to use is completed passes. If x1 is a column listing the number of passes completed in each game in 2006 in random order and x 2 is a column listing the number of passes completed in each game in 2007 in random order and you consider these as independent random samples and wish to compare the mean passes completed per game and to get results that verify your statement, what are your hypotheses? What is your method? H : 2 H : 2 0 Solution: If this is a valid method for testing for improvement. 0 1 or 0 1 . If H 1 : 1 2 H 1 : 1 2 0 H 0 : D 0 . If you have decided to use means, you must believe that the Normal D 1 2 , then H 1 : D 0 distribution applies. The total sample size is too small to use Method D1, which means that D2 or D3 should work. You could test the variances for equality and use D2, or not bother and use D3. You may want to see the methodology below. If we use a confidence interval it will be of the form D d t s d . If we use a test ratio, we will compare t d D0 with t . sd If we use a critical value for d , we will use d cv D0 t s d . b. Your roommate, Gigglebutz, says that you have done this all wrong, because the distributions have not been shown to be Normal. You do the test described in section 6.3 of the text and sadly agree with Gigglebutz that the distributions are probably not Normal. What is your test and method now? Solution: If the sample is small and we have no reason to believe that a symmetrical distribution applies, H 0 : 1 2 we should compare medians. If is the median . Since we are comparing medians and we H 1 : 1 2 still think that the data are independent random samples, use Method D5b. c. Gigglebutz is still not satisfied and points out that these are not independent random samples and what you should do is rearrange all your data so that each column represents a year and each row an opponent. Any school that was not played in both years should be dropped from your data. What is your test and method now? Solution: If the sample is small and we have no reason to believe that a symmetrical distribution applies, H 0 : 1 2 we should compare medians. If is the median. . Since we are comparing medians and we H 1 : 1 2 have just put the data into pairs, use Method D5b. 2. Standard deviation is often a measure of reliability. A manufacturer is providing a connector and is getting complaints that the connector is often too large or too small for the intended use. A sample of 250 connectors produced by the process in current use is taken. Then a new process is tried and a sample of 45 connectors is taken. Find the hypotheses and method to show that the new process produces a more reliable product. Solution: When you see words like reliability or variability, think variance or standard deviation. The statement that a new process is more reliable means that the variance or standard deviation of the new process is smaller. Since this statement does not include an equality, it must be an alternate hypothesis. 4 252solngr3-061 11/01/06 H 0 : H 0 : 12 22 2 2 1 2 or . In terms of the variance ratio 12 or 22 , the alternate hypothesis 2 1 H 1 : 12 22 H 1 : 1 2 rules, so H 1 : 12 22 1 . This means that the null hypothesis is H 0 : variances, use Method D7. Compare the ratio s12 s 22 12 22 1 . Since you are comparing against F . 3. (Dummeldinger pg 148) We are interested in attitudes about sexual discrimination. Various groups were asked if they believed sexual discrimination is a problem in the United States. .05 a. Out of 50 men 11 believed that sexual discrimination is a problem. Out of 40 women 19 said that sexual discrimination is a problem. Is there a significant difference between male and female attitudes? Solution: The only possible measure that can come out of this statement is the proportions of men and women that believe sexual discrimination is a problem. ‘Difference’ means not equal. p x1 1 n1 H 0 : p1 p 2 H 0 : p1 p 2 0 H : p 0 If or . If p p1 p 2 , then 0 . Since we are H : p p H : p p 0 x 2 2 1 1 1 1 H 1 : p 0 p2 2 n2 comparing proportions from independent samples, use Method D6a. If we use a confidence interval it will be of the form p p z s p . 2 If we use a test ratio, we will compare t d D0 with z . sd 2 If we use a critical value for p , we will use pcv p0 z p . 2 b. A group of 100 men are picked from upper levels of a group of large American firms. They are asked if sexual discrimination is a problem in the United States. 38 said ‘yes.’ They are then sent to a program on mentoring female executives where they discuss, among other things, what problems seem to be peculiar to young female executives. Afterwards the men are asked the same question. This time 41 said ‘yes.’ Assuming that these 100 men are representative of all executives who are trained to mentor female executives, has the training increased the proportion that believes there is a problem? Solution: It clearly says that we are comparing proportions and that we want to see if p 2 p1 . Since this p x1 1 n1 H 0 : p1 p 2 does not contain an equality, it must be an alternate hypothesis. If we can use H 1 : p1 p 2 p2 x2 n2 H 0 : p1 p 2 0 H 0 : p 0 or . If p p1 p 2 , then . We do not have independent samples but are H 1 : p1 p 2 0 H 1 : p 0 sampling twice from the same sample. Since we are not comparing proportions from independent samples, question 2 question 1 yes no use Method D6b. Note that our setup is , We know that x11 x12 x1 38 and yes x11 x12 x no 21 x 22 that x11 x 21 x 2 41 , but we do not know x 21 or x12 . Since these numbers are missing, we will have to go back to our original data and compute them. x x 21 We will then compare z 12 against z . x12 x 21 5 252solngr3-061 11/01/06 c. Random samples (each 100 people) of male 1) high school graduates, 2) college graduates and 3) people with MBAs are asked the same question. Out of the first group 35 said ‘yes.’ Out of the second group 38 said ‘yes.’ Out of the third group 39 said ‘yes.’ Is there a significant difference between the attitudes of the three groups? p x1 1 n1 H 0 : p1 p 2 p 3 x Solution: If p 2 2 . This is a chi-squared test of homogeneity. Since we are n 2 H : not all ps equal. 1 p x3 3 n3 comparing multiple proportions, use a chi-squared test. The O that we need must account for all of each O HSGrads CollegeGr MBAs 35 Yes 38 39 sample and is . No 62 61 65 Extra credit is at the end of the assignment. 4. (Dummeldinger) A researcher took a random sample of n graduates of MBA programs, which included n1 women and n 2 men. Their starting salaries were recorded. Use 1 and/or 1 for population parameters for women and 2 and/or 2 for men. Choose hypotheses and methods for the following. .10 a. n1 n 2 150 . The researcher wants to show that men have a higher mean starting salary than women. H 0 : 1 2 Solution: If this is a valid method for testing for a difference in starting salaries, use or H 1 : 1 2 H 0 : 1 2 0 H 0 : D 0 . If D 1 2 , then . If you have decided to use means, you must believe H 1 : 1 2 0 H 1 : D 0 that the Normal distribution applies. The total sample size is large enough for Method D1, which means that D2 or D3 should work as well. You could test the variances for equality and use D2, or not bother and use D3. b. n1 18 and n2 12 . The researcher wants to compare the dispersion of men’s and women’s salaries. The researcher has no prior opinion as to which is more variable. Solution: When you see words like reliability or variability, think variance or standard deviation. Since the research has no idea of which dispersion is larger, this is a two-sided test of equality of variances unless 2 2 H 0 : 1 2 there is a good reason to believe that the Normal distribution does not apply. We can use H 1 : 12 22 H 0 : 2 22 1 2 or . In terms of the variance ratio 12 or , it doesn’t matter which ratio we use, so we 2 12 H 1 : 1 2 have H 0 : 12 22 1 and H 1 : 12 22 1 or H 0 : 22 12 1 and H 1 : 22 12 1 . Since you are comparing variances, 6 252solngr3-061 11/01/06 use Method D7. In practice this means to compare the ratio ratio s 22 s12 s12 s 22 n 1, n 1 against F 1 2 and to compare the 2 n 1, n 1 against F 2 1 . 2 c. n1 18 and n2 12 and the null hypothesis in 4b) has not been rejected. Dummeldinger gives the following data. x1 48266 .70 s1 13577 .63 x 2 55000 .0 and s 2 11741 .25 . The researcher wants to show that men have a higher mean starting salary than women. H : 2 Solution: If this is a valid method for testing for a difference in starting salaries, use 0 1 or H 1 : 1 2 H 0 : 1 2 0 H : D 0 . If D 1 2 , then 0 . If you have decided to use means, you must believe H : 0 2 1 1 H 1 : D 0 that the Normal distribution applies. The total sample size large is not large enough for Method D1, which means that D2 or D3 should work. But you have tested the variances for equality and shown that you could use method D2. Extra credit is at the end of the assignment. 5. (Dummeldinger) In order to find the effect of membership in a fraternity or sorority on grades, 250 students who had joined fraternities or sororities were picked as a random sample. a. GPA’s were recorded for each student for the semester before ( x1 column) and the semester after they joined ( x 2 column). Mean GPAs were compared. Choose hypotheses and method. H 0 : 1 2 Solution: If this is a valid method for testing for a difference in starting salaries, use or H 1 : 1 2 H 0 : 1 2 0 H 0 : D 0 . If D 1 2 , then . If you have decided to use means, you must believe H 1 : 1 2 0 H 1 : D 0 that the Normal distribution applies. The total sample size is large enough for Method D1, which means that D2 or D3 should work as well. You could test the variances for equality and use D2, or not bother and use D3. b. Dummeldinger says that a 95% confidence interval for the difference between the means was 0.24 to 1.02. What is your conclusion? Solution: The confidence interval has been thrown at us, so we must assume that the author knew what he was doing. Since the interval includes zero and a 2-sided test was appropriate if the hypotheses are as in 5a), we can say that there was no significant difference between the before and after means. Extra Credit Problems 3d. (Dummeldinger pg 148) We are interested in attitudes about sexual discrimination. Various groups were asked if they believed sexual discrimination is a problem in the United States. .05 a. Out of 50 men 11 believed that sexual discrimination is a problem. Out of 40 women 19 said that sexual discrimination is a problem. Is there a significant difference between male and female attitudes? b. A group of 100 men are picked from upper levels of a group of large American firms. They are asked if sexual discrimination is a problem in the United States. 38 said ‘yes.’ They are then sent to a program on mentoring female executives where they discuss, among other things, what problems seem to be peculiar to young female executives. Afterwards the men are asked the same question. This time 41 said 7 252solngr3-061 11/01/06 ‘yes.’ Assuming that these 100 men are representative of all executives who are trained to mentor female executives, has the training increased the proportion that believes there is a problem? c. Random samples (each 100 people) of male 1) high school graduates, 2) college graduates and 3) people with MBAs are asked the same question. Out of the first group 35 said ‘yes.’ Out of the second group 38 said ‘yes.’ Out of the third group 39 said ‘yes.’ Is there a significant difference between the attitudes of the three groups? d. You have enough information to do two of the three problems above. What additional information do you need? Solution: In 3e) I will demonstrate that we have enough information to do 3a) and 3c). If you look at the solution to 3b) above you will notice that I complained that we do not know x 21 or x12 . These represent people who have changed their answers between Question 1 and Question 2. 3e. Do one of the three problems. State your conclusion clearly. .05 Solution: We cannot do 3b. If we wish to finish 3a), remember that x1 11 , n1 50 , x 2 19 and n 2 40 . The proportions are 11 19 .2200 and p 2 .4750 . q1 1 p1 1 .2200 .7800 and 50 40 q 2 1 p 2 1 .4750 .5250 . p p1 p2 .2200 .4750 .255 . The Formula Table says the following. Interval for Confidence Hypotheses Test Ratio Critical Value Interval pcv p0 z 2 p Difference p p 0 H 0 : p p 0 p p z 2 s p z between If p 0 p H 1 : p p 0 p p1 p 2 proportions 1 1 If p 0 p p 0 q 0 p 0 p 01 p 02 p1 q1 p 2 q 2 q 1 p n1 n 2 s p p 01q 01 p 02 q 02 p or p 0 0 n1 n2 n1 n2 n p n2 p 2 p0 1 1 n1 n 2 Or use s p p1 H 0 : p1 p 2 H 0 : p1 p 2 0 H : p 0 Our hypotheses are or or if p p1 p 2 , 0 . H 1 : p1 p 2 H 1 : p1 p 2 0 H 1 : p 0 x x 11 19 30 n p n p 50 .2200 40 .4750 0.3333 1 1 2 2 z 2 z.025 1.960 . p0 1 2 n1 n2 50 40 90 n1 n2 50 40 q0 1 p0 1 .3333 .6667 p p 0 q 0 1 n1 1 n3 .3333 .6667 150 1 40 .22222 .02 .025 .0099999 .1000 (Only one of the following methods is needed!)Test Ratio: z p p 0 p .255 0 2.55 . Make a .1000 diagram showing a 95% 'accept' region between -1.960 and +1.960. Shade the 'reject regions below -1.960 and above 1.960. Since -2.55 lies in the lower 'reject' region, reject H 0 . or Critical Value: pcv p0 z p 0 1.9600.1000 0.196 . Make a diagram showing a 95% 2 'accept' region between -0.196 and +0.196. Shade the 'reject' regions below -0.196 and above 0.196. Since p .255 lies in the lower 'reject' region, reject H 0 . p1 .2200 , p 2 .4750 , q1 .7800 and q 2 .5250 . 8 252solngr3-061 11/01/06 or Confidence Interval: p p z sp 2 s p p1 .2200 , p 2 .4750 , q1 .7800 and q 2 .5250 . p1 q1 p 2 q 2 .2200 .7800 .4750 .5250 .003432 .006234 .0096664 0.0983 . n1 n2 50 40 So p .255 1.960 0.0983 .255 .193 or -.448 to -.062. Note that this interval does not include zero and thus causes us to reject the null hypothesis. Minitab output follows. MTB > PTwo 50 11 40 19. #Format of this instruction is n1 , x1 , n 2 , x 2 . Test and CI for Two Proportions Sample X N Sample p 1 11 50 0.220000 2 19 40 0.475000 Difference = p (1) - p (2) Estimate for difference: # Just prints out the two proportions. -0.255 #Prints p p1 p2 .2200 .4750 .255 95% CI for difference: (-0.447699, -0.0623008) #Exactly the same as the interval above. Test for difference = 0 (vs not = 0): Z = -2.59 P-Value = 0.009 #The z was evidently computed using s p instead of p . Bad! H : p p 2 p3 To finish 3c) note that we had 0 1 . The O that we found accounted for all of each H 1 : not all ps equal. sample and was, with row and column totals, as below. O HSGrads CollegeGr MBAs Total pr 35 Yes 38 39 112 .3733 . Row proportions are gotten by No 62 61 188 .6267 65 Total 100 100 100 300 1.0000 112 dividing row totals into the overall total, for example .3733 . We now get our Expected E table by 300 using the row proportions to multiply the column totals. For example we replace 35 by .3733 100 37 .33 . E HSGrads CollegeGr MBAs Total pr Yes 37.33 37.33 37.33 112 .3733 The expected array is No 62.67 62.67 62.67 188 .6267 Total 100 100 100 300 1.0000 The formula for the chi-squared statistic is 2 O E 2 or 2 E formulas are shown below. DF r 1c 1 2 13 1 2 . index O E 1 35 37.33 2 65 62.67 3 38 37.33 4 62 62.67 5 39 37.33 6 61 62.67 Total 300 300.00 So we have 2 O E 2 2 2 E E O2 E O -2.33 2.33 0.67 -0.67 1.67 -1.67 0.00 0.3705 or 2 O2 n . Both of these two E E O 2 E 0.145430 32.8154 0.086627 67.4166 0.012025 38.6820 0.007163 61.3372 0.074709 40.7447 0.044501 59.3745 0.370456 300.3704 O2 n 300 .3704 300 0.3704 . If we compare E our results with .05 5.9915 , since our computed value of chi-squared is less than the table value, we do not reject our null hypothesis. Minitab output for the same problem follows. 9 252solngr3-061 11/01/06 MTB > print c11-c13 Data Display Row 1 2 O1 35 65 O2 38 62 #Columns are in c11 – c13 and are labeled O1-O3. O3 39 61 MTB > ChiSquare C11-C13. Chi-Square Test: O1, O2, O3 Expected counts are printed below observed counts Chi-Square contributions are printed below expected counts O1 35 37.33 0.146 O2 38 37.33 0.012 O3 39 37.33 0.074 Total 112 2 65 62.67 0.087 62 62.67 0.007 61 62.67 0.044 188 Total 100 100 100 300 1 #You should be able to find the elements of E, O and E O 2 E in the #table above. Chi-Sq = 0.370, DF = 2, P-Value = 0.831 4d. Do problem 4c. (Hint: before you start, move the decimal point so that the sample means and standard deviations are in thousands.) State your conclusion clearly. H : 2 H : 2 0 Solution: We already have our hypotheses 0 1 or 0 1 or if D 1 2 , H 1 : 1 2 H 1 : 1 2 0 H 0 : H 0 : 12 22 1 2 or . Finally we have 2 2 H 1 : 1 2 H 1 : 1 2 x1 48266 .70 x 2 55000 .0 and . All of these are to be expressed in thousands. n1 18 , n2 12 , s1 13577 .63 s 2 11741 .25 H 0 : D 0 . We know that we cannot reject H 1 : D 0 The formula table gives us the formulas below. Interval for Confidence Hypotheses Interval Difference H 0 : D D0 * D d t 2 s d between Two H 1 : D D0 , 1 1 Means ( sd s p D 1 2 n1 n2 unknown, variances assumed equal) DF n1 n2 2 Test Ratio t sˆ 2p d D0 sd Critical Value d cv D0 t 2 s d n1 1s12 n2 1s22 n1 n2 2 DF n1 1 n 2 1 17 11 28 n1 n 2 2. d x1 x 2 48.26670 55.00000 6.7333 . sˆ 2p n1 1s12 n2 1s 22 n1 n 2 2 17 13.57763 2 1111.74125 2 28 166 .0861 . This is the pooled variance. sˆ p 12.8874 s d sˆ p 3133 .984619 1516 .426467 28 .10 so t.28 10 1.313 . 1 1 1 1 1 1 166 .0861 166 .0861 .0555555 .0833333 sˆ 2p 18 12 n1 n 2 n n 2 1 23.0675 4.8029 . Recall that our alternate hypothesis is H 1 : D 0 so this is a left-sided test. 10 252solngr3-061 11/01/06 (Only one of the following methods is needed!) Test Ratio: t If this test ratio lies below t .28 10 1.313 , reject H 0 . t x x 2 10 20 d D0 . or t 1 sd sd d D0 6.7333 0 1.401 . Make a sd 4.8029 diagram with zero in the middle showing a shaded ‘reject’ region below -1.313. Since -1.401 falls in the 28 28 'reject' region, reject H 0 . Or you can say that, since 1.401 falls between t .10 1.313 and t .05 1.701 , for a one-sided test, .05 p value .10 . Since the p-value is below .10 , reject H 0 . Critical Value: d CV D0 t 2 s d or x1 x 2 CV 10 20 t 2 s d . For a left-sided test we want a critical value below D0 0. If d x1 x 2 is below the critical value, reject H 0 . d CV D0 t s d 0 1.313 4.8029 6.3062 . Make a diagram with 0 in the middle showing a shaded 'reject' region below -6.3062. Since d 6.7333 falls in the 'reject' region, reject H 0 . Confidence Interval: D d t 2 s d becomes D d t s d for a left sided test. We already know that t s d 1.3134.8029 6.3062 so we can say D 6.7333 6.3062 or D 0.4271 . Make a diagram with -6.7333 in the middle. Represent the confidence interval by shading the area below -0.4271. Since zero is not in this area, reject H 0 . Or simply note that it is impossible for D 0 as stated by the null hypothesis and D 0.4271 . The Minitab run gives us the following. The instruction TwoT is followed by the size, mean and standard deviation of sample 1 followed by the size, mean and standard deviation of sample 2. The ‘alternative -1’ is always used to set a left sided test. Note that the pooled standard deviation and the value of t are identical to the values that I got using my calculator. MTB > TwoT 18 48266.70 13577.63 12 55000.00 11741.25; SUBC> Pooled; SUBC> Alternative -1. Two-Sample T-Test and CI Sample N Mean StDev SE Mean 1 18 48267 13578 3200 2 12 55000 11741 3389 Difference = mu (1) - mu (2) Estimate for difference: -6733.30 95% upper bound for difference: 1437.00 T-Test of difference = 0 (vs <): T-Value = -1.40 Both use Pooled StDev = 12887.4400 P-Value = 0.086 DF = 28 11