Unit 6 Hypothesis Testing 6.1 What you will learn in this Unit Statistical Inference is concerned with making conclusions about the unknown state of nature. When we fomulate a model we are expressing our ‘feelings’ or suspicion about what we think could be true. ‘That is we make conjectures or guesses that may or may not be true. Hypothesis testing, as the name implies, is concerned with checking whether or not we should believe our guesses strongly,.doubtfully or not all. As for confidence intervals these hypotheses could be about means, variances, proportions etc. ‘This Unit is devoted testing of statistical hypotheses. By the end of this Unit you are expected to be able to 1. define a statistical hypothesis 2. distinquish between null and alternative hypotheses, 3. define a critical or rejection region, 4, define type I and type II errors 5. define significance level or size of a test, 6. define the power of a test, 7. explain how the p-value is calculated and how it is used to make a statistical decision, 8. test hypotheses concerning the mean, variance, proportion ‘two means, two variances and two proprtions. 86.2 Statistical Hypotheses ‘The object of experimental research is to gain some insights or knowledge about levels, patterns, trends, etc. of various phenomena, physical, soci-economic, etc. For example, suppose one wants to compare a new method of cultivation with a standard one. Sup- pose also that if is known from past experience that the mean yield with a standard or old method is 4 = 100 units per acre. To test: whether a new cultivation method increases or reduces the yield we must formulate an appropriate statistical hypthesis and look for sufficient experimental ie., statistical evidence to support or refute our hypothesis. But what exactly do we mean by a statistical hypothesis? W Definition 6.1 A statistical hypothesis is a belief or suspicion about the parameter(s) of a population or statistical model for a data generating process or system. yDefinition 6.2 The null hypothesis, usually denoted by Ho is a statistical hipothesis which asserts or says that no change from the existing state or level has occurred. “Definition 6.3 The research hypothesis, commonly referred to as the alternative hy- pothesis, is the hypothesis which declares what a researcher believes or suspects to be the new or, true but previously unknown, state of nature. Voges A statistical tes OPHiyppthesis is a procedure or rule which is used in conjuction with experimental data to decide whether or not the null hypothesis should be rejected in farvour of the research or alternative hypothesis, Definition 6.4 A test statistic T is a calculation from the experimental data whose value is used to decide whether or not Ho should be rejected. Definition 6.5 A critical region or rejection region of a test using a test statistic T, is a setC chosen so that if the value of T falls in it, then Ho is rejected, and hence the term rejection region. As for confidence intervals, we shall be concerned mostly with tests involving data from normal distribtutions. However to illustrate a number of concepts, we shall begin the discussion with some relatively simple examples. We first note that in a statistical test the form of the alternative hypothesis and hence the critical region, dependes on our suspicion. ‘A test of the hypothesis Ho: =O against Hy: 0% is said tobe‘ (i) one sided and right tailed if Hy :8 > 0 (ii) one sided and left tailed if Hy : 0 < (ii) two sided if Hy: @ # 65 79‘A onesided test is one that locates the critical region on one side of the sampling distribution involved while a two-sided test is one that locates the critical region on both sides of the distribution. ‘A statistical hypothesis can be rejected on the basis of experimental evidence when in actual fact it is correct, or accepted when, in fact, it is false. Thus in making a statistical decisions one must be careful not to make these errors. We discuss these errors and their probabilities below. Activity 6.1 1. Explain what you understand by the term ‘hypothesis’. | 2. Distinguish between null hypothesis and alternative hyy esis. | 6.3 Type I and Type II Errors ; so made is called type I error and the probability of this error is called the significance Definition 6.6 In statistical test, if Ho is rejected when in fact it is correct, the error level or size of the test and is commonly denoted by a i.e. a=P(TE Cle) |) Re C cecktek Maem Definition 6.7 Ina statistical test if Ho is retained i.e. not rejected when in fact, it is false s0 that Hy is correct, the error so made is called type I error. The probability if this error is usually denoted by B and is given by 6 = P(E Alm) where A=C' is the acceptance region which is the complement of C. y Example 6-1 Let X1 be a random sample of size 1 from the distribution oF, 2>0, 06 {1,2}. and suppose that one wants to test the hypothesis Ho: =1 against Hy: using X; as the test statistic and C = [1.5,00) = 1.5 < 2 < 00 as the rejection region. Caleulate (i) the probability of type T error i.e the significance level of the test. (ii) the probalility of type IT error. 80Solution 6.1 (i) The protabilty of type I error is 7 a = P(C\to) P(X > 1.5]H0) 4 P(X, > 1.5)0=1) = f de hs ys = 0.2231 (ii) The probabilty of type IT error is B = P(A) P(X: < 1.510) \ = P(X, <15)0=2) 15 = f ze~*dz, since under Hy @=2 =Lbe 15 $1 5, integrating by parts 1-2.5e715 q 1 - 0.5578 | = 0.4422 Activity 6.2 1. Define the term ‘significance level’. 2. Distinguish between type I and type II errors 6.4 Power of a Statistical Test Definition 6.8 The power of « statistical test with eritical region C is defined a» P(C\ih) = 1-P(A\h)=1-8 ‘The power of a test represents the probability of correctly rejecting Ho ie. rejecting Ho when it is indeed false and Hy is true. Example 6.2 In the preceding ezample, calculate the power of the test? Solution 6.2 The probability of correctly rejecting Ho when @ = 2 is P(C\Eh) = P(X 2 1.5\Hi) P(CI) = P(X 2 15/0=2) = [sede hs = 25e1F = 0.5578. 81Wf Example 6.3 Let Xi,...,Xs be a random sample from the distribution f(@,8) = 02°" O0 and suppose that one wants to test the hypothesis Ho:6 1 against Hy :0<1 using T = Xi) = max{X;,...,X5} as the test statistic and the interval C= (0,0.5] as the critical region. (i) Determine the significance level of the test and interpret this value. (ii) Determine the critical region with significance level « = 0.05. Solution 6.3 If Ho is true then the underlying distribution is uniform with density S (2) =1 and distribution function F(e) = 2. (a) It follows from the eampling distribution of the mazimum that Y = Xiq) has density fy (y) = Sy". Thus the significance level of the test is @ = P(X») < 0.5/8 = 1) os = Sv(y)dy 5 8 fo otey lo = Bp fo.s}§ 0.0313. (b) Let (0, ¢] be the critical region of size a = .05 for some c between 0 and 1. Then Hence ¢ = {0.05}? Example 6.4 Suppose that a coin with probability p of getting a head is tossed 5 times. The coin is considered to be fair if the number of heads X € {0,1,}U {4,5}. (Y) State clearly and in symbols the null and alternative hypotheses for this problem. (ii) State the critical region for the test. (iii) Find the probability of type I error? 82(iv) If the probability f a head is 0.6, ealeulate the probability of type IT error. Hence determine the power of the test. Solution 6.4 Let p be the probability of getting a head and X be the number of hends in 5 tosses of the coin. Then (3) the null and alternative hypotheses are Ho :p=0.5 and Hi # 0.5, two-sided. (i) The critical region is C= {0,1} UL4,5}- (iii) Under Ho : p = 0.5 X is @ binomial random variable with paramters (5,0.5) Hence the probability of type I error is a = P(C\Ho) P(C\p = 0.5) P(X =0) + P(X = 1) + P(X =4) + P(X = 5) = (Jestest + () (0.5)'(0.5)* + (') (0.5)*(0.5) + () (0.5)9(0.5)° = 0.374 (i) Under Hy : p = 0.6 X is a binomial random variable with paramters (5,0.6)- Hence 8 (Alf) (Alp = 0.6) P(X =2)+ P(X =3 -(° 2 3. 5 3 2 () (0.6)2(0.4)° + ( (0.6)%(0.4) 0.576. Hence the power of the test is 1 — B=1-0.576 424, We now turn to more practical tests of statistical hypotheses in which the data are assumed to come from the normal distribution. 83The general form of the test statistic for testing the hypothesis Ho :0 =, concerning a population parameter @ is given by the so-called pivotal statistic estimate - true value 8 std of extimate % where og is the standard error i.e. standard deviation of the estimate. Most of the test statistics considered here, except in cases involving variance, will be of the above form. A complete list of the test statistic is given at the end of this chapter. 6.5 Tests Concerning the Population Mean y. Let X1,Xz,..-,Xp be a random sample from a normal population (y, 07). Then the | test statistic for testing the hypothesis Ho” is given by () Z = 728 if 0 is known (i) T= 27% if 07 is not known, ‘The corresponding critical regions or cut-off values depend on the sampling distribution of the test statistic involved, the form of alternative hypothesis H; as well as the significance level at which the test is being conducted. In the second year course on inference you will learn how to derive the appropriate critical regions. In this course presumably good guide is that if the alternative hypothesis is tailed to the right | then the critical region is also tailed to the right i.e. of the form[e,oo). A similiar interpretation holds for the other forms of Hy. The critical values i.e. end-points of the critical region, are chosen so that when Ho is true the probability of the test statistic falling in the critical region is equal to the significance level a. The following table shows the appropriate critical regions > corresponding to each of the three forms that H; can take and a diagram to illustrate the regions in the distribution being used. 84(2) ' alternative critical region Ay: B> Ho [eas 3) _7 | be 4 Hi: p< bo (-00, za) ef, | = “tO F(a) meiptpe (Commas) Figure 6.1: One sided and two sided tests From the diagram it is clear that for a one sided and right tailed tes, the critical value ¢= %q must be chosen so that X = Ho = (ee «)=9 In this case, the null hypothesis should be rejected in favour of Hy for large values of Z, that is when an ‘observed value z of Z exceeds ¢ = Za- “Another criterion for rejecting Ho is based on ® value called the p-value or probability an theinich is the least significance level at which @ “statistical test is significant. You will learn more about, this rejection ct ‘iteria in your second course on Inference. 85 ————————Example 6.5 A standardized placement test has been given for the past couple of years to prospective freshmen(first year college students). A mean mark of » = 80 with @ standard deviation of ¢ = 4 has been observed over this period, Due to changes in the high school curriculum it is suspected or believed that the mean score on this test has changed with the standard deviation remaining constant. Suppose that the test is given to a group of 16 students and average mark of 82 is obtained. Test the hypothesis that the mean score has increased, stating clearly and in symbols, the null and alternative hypotheses. Use a 5% significance level and assume that the data are normally distributed. Solution 6.5 Let 4 be the new population mean of the test scores. Assuming that the current test scores follow a normal distribution with unkown mean js and known standard deviation o = 4, the testing procedure is as follows. 4. Hypotheses: Hy: 1 =80 against Hy: p> 80 . Critical region: 2 > 1.645 = (1.645, 00), with za = 2.95 = 1.645. 2 3. Stati = S5ff = 2 € [1.645,00) 4. Decision: Ho is rejected since the test statistic is in the rejection region. 5 . Conclusion: Since Ho is rejected we conclude that the there is sufficient statistical evidence to believe that the population mean of the test scores has increased. ‘Example 6.6 Suppose a random sample of size 20 from a normal ‘population N (4,02) has mean 52.95 and standard deviation 16.99. Test the hypothesis Hoi Use a 5% significance level. 4 10 against Hy: 60 Solution 6.6 Since the population variance is not known the table value for a two-sided test with a =.05 from the t-distribution, is ty_1,0/2 =th9,o.025 = 2.09. Thus, Critical region is t < —2.09 and t > 2.09= (—c0, ~2.09] U [2.09, 00). Test Statistic is T = Foe = 295-60 — 1.856, The statistical decision ts to retain Ho and conclude that there is no sufficient evidence to believe that u #60. If the sample size is large then using the t-distribution is equivalent to using the normal distribution. We illustrate this in the next example. Example 6.7 In packing sugar certain bags should weigh thg. To see if the manufac- turer puts enough sugar in the bags, the contents of random sample of 100 such bags were weighed. The average contents turned out to be 0.99 kgs with a standard deviation of 0.06. State the null and research hypotheses for this problem. Hence test at the 0.05 significance level, whether or not the manufacturer satisfies the requirement,Lot be the true mean weight of a bag. Then Hypotheses are: Ho: = kg, versus Hiya < 1 Statiatie is Z = ofis = aseaios = 187 Critical region is ?-< ~1.645 = (~00, ~ 1.645] with 2008 645. Decision: Since the test statistic isin the critical region Ho js rejected and the conclusion eithat the manufacturer does not put enough sugar in the bags. \ . wrActivity 6.4 $1. Suppose that the observations 24, = 1.6, 38, 4.0, ~5.2, —1.5, 2.0, 08, 32 are independent and identically distributed as N(s,0%). Test the hypothesis that the mean of the population is 1.2. ~ 2. Suppose that the number of customers entering a eoffee-shop i day, follows at ‘eae pprozimately, the Poisson distribution with mean 60: By advertizing (his coffee) in a local newspaper the owner of the shop hopes to increase the number of his customers. (i) Write down the appropriate null and alternative hypotheses. (ii) If the number of customers on 90 randomiy selected days averages to 65 use te Contral Limit theorem to test the hypotheses in part (). Take a = 05. (iii) Calulate the power of the test and relate it to the probability of a possible error in your conclusion. 6.6 Tests Concerning the Variance From the sampling distribution of the (scaled) sample variance which is a chi-square sreumtbation with n ~ 1 degrees of freedom we readily derive a test concerning the population variance. We give an example to that effect Pelow. Y Example 6.8 A manufacturer of computer batteries claims that his batteriés wil last, ‘on the average, 3 years with a standard deviaion of 1 year. Suppose ‘that the life times Of 5 auch batteries have the following lifetimes 3.0, 1.9, 24, 4.2, 3.5 Do you believe that the lifetimes of these batteries have @ standard deviation of 1 year? Use a 5% significance level. Solution 6.7 The sample mean and sample variance are, respetively, # = 3.0 and s? =0.815. 1. Hypotheses: Hoo? =1 against, Ho:a? #1 2, Critical regio: x? < X4,o.ars = 0.484 and x? > x40.028 N1.143= (0, 0.484] U (11.143, 00) a 8, Statistic: x2 = “DE 0318) M0818) — 3.26 87 — rs4. Decision: Since the test statistic is in the acceptance region, we do not reject Hy. 5. The conclusion is that the standard deviation is 1 as claimed. Activity 6.5 Consider the data and assumptions of the preceding ezample. Test the hypothesis that the population variance is greater than 1.44. Use a significance level of a= 90%. 4 6.7 Tests Concerning a Propotion p ¥ Suppose it is desired to test the hypotheses Hoi p= po versus Hi: p= po Then the test statistic is where # is the sample estimate of p. Example 6.9 A manufacturing company submitted a claim that leas than 10% of its products are defective. Some consumers of the products doubt this claim, In an ex- periment to test the validity of the claim, 100 items randomly selected from a large lot of items were inspected and & were found to be defective. Is this sufficient evidence to confirm the manufacturers claim? Solution 6.8 let p be the true propotion of defectives. Then Ho:p=0.10 against Hy : p< 0.10 -167 Test Statistic is % Critical region is Z < —20.05 = (00, ~1.645] Conclusion: We reject Hy and conclude that the proportion of defectives is indeed less. than 10%. , It is important to observe that from the consumers’s point of view, the proportion of defectives is at least 10%. Activity 6.6 It is believed that the proportion of female students at the University of Zimbabwe is less than 0.5. To investigate this view a random sample of 200 students was taken and the number of female students counted. (i) State clearly and in symbols, the null and alternative hypotheses for this problem. (ii) Find an appropriate, critical region with real significance level = 0.058. (3ii) Suppose that 60 of the 200 students were female. What is your conelusion to the above test? (Go) If the true proportion of female students is 0.4, calculate the probability of a type Tl error. Hence or otherwise determine the power of the test, 88% 6.8 ‘Testing for the Difference Between Two Means ‘The sampling distributions used here are exactly the same as those used in constructing confidence intervals. MM Example 6.10 Suppose two independent samples from N(u1,0}) and N(b2,03 yield the following summary satistics: sample n @ 8 my 10 4.2 138 2 1 84 641 Assuming that the population variances are equal, test the hypotheses 1 = Be Ho: #1 =be versus Hy: Use a = 0.05 significance level. Solution 6.9 Assuming equal population variances, the pooled estimate of the common standard deviation is ocr 5)? + (6.11)? + Hence the test statistic is T= 2172. = 42-34 = 0.235, The rejection region is t < —ths.0.028 = 2.13 and t > 2.13.” Thus, Ho is not rejected. 5 69. ‘Activity 6.7 Suppose two independent samples from N (1,02) and N (12,03 yield the following summary satistics: . sample n = 8 m1 20 24 1.84 2 25 2.6 2.00 Assuming that the population variances are equal, test the hypotheses Ho: =y2 versus Hy tpn # ta Use «= 0.05 significance level. 6.9 Testing for Equal Population Variances In practice we can not just assume that the population variances are equal. One would normally test this assumption first. If the hypothesis of equal variances is not rejected then that would justify the procedure based on the assumption of equal variances. If the test is significant, ie. if Ho is rejected in favour of unequal population variances, then one would proceed to use an appropriate testing procedure that does not require population variances to be equal. To test for equality of population two variances we use the sampling distribution of the ratio of sample variances. 89 ml# pxample 6.11 In the above ecample, on two samples, test the hypothesis that the population variances are equal. Use a 10% significance level Solution 6.10 1. Hypotheses: Ho:o?=0} against Ho:0} #03 f(t) _ 1398/9 3, Statistics Sm) = Fle = 0.9726. 2. Critical region: F < Finsm)a-al2 and F > Fnyng)c/2 = (0; Flee,o.98) U [F(9.6),005: 09) = (0; 0.297] U [4.10, 00). 4. Decision: Since the test statistic is in the acceptance region, we do not reject Ho. 5. The conclusion is that the population variances are not significantly different. AC ‘ity 6.8 Consider the data in the prvious example. Test the hypthesis that the population variances equal, State your conclusions at the 95% significance level. ‘The following example illustrates the large sample case where a test of equality of two means can be performed when the population variance are unequal. Example 6.12 A manufacturer of light bulbs wishes to determine whether a new method of seang bullbs will increase ‘the life times of his bulbs. A random sample sample of size 49 bulbs sealed using the old method produced a mean lifetime of 6.0 months with a standard deviation of 0.5 months. A second random sample sample of size 36 of bulbs sealed using the new method yeilded a mean lifetime of 6.5 months and a standard deviation of 0.9. State the null and research (alternative) hypotheses for this problem, Can we conelude with a 95% probability of being correct that the new method does increase the lifetime of a bulb? Solution 6.11 Let p be the mean lifetime of bulbs sealed using the old method and ng be the mean lifetime of bulbs sealed using the new method. Then hypotheses to be tested are Ho: jy = fla mbonversus Hy: p2 > 1 Critical region is Z > 29.05 = 1.645 Statistic is Z .65. The statistical a decision is to reject Ho and conclude that the new method increases the lifetimes of the bulbes. cems superior to. Activity 6.9 Suppose two independent samples from N(y1,0%) and N (ja, 0} yield the {following summary satistics: sample n 2 8 a 20 24 1.84 t. 25 26 2.00 Test the hypothesis that the variances are equal. Use a = 0.05 significance level. 90——$— $e 6.10 ‘Testing for Difference Between two Means for Paired Samples fin this section we illustrate how to test for the difference between two means when the observations in the two samples are paired or matched. Example 6.13 Five ‘samples’ of a chemical substance were used to determine if there is a difference between a chemical analysis and an X-ray analysis of iron content. Heach sample was split into two sub-samples and the two types of analysis performed. The data are displayed in the table below. sample Analysis 1 2 3 4 65 X-ray = 2.0 2.0 28 21 24 Chedmical 22 1.9 25 23 24 Assuming the populations to be normal, test, at the 5% significance level, whether the ‘two method of analysis produce different iron contents. Solution 6.12 Let yi be the mean iron content produced by chemical method, pz be the mean iron content produced by X-ray analysis and jug = 1 — pa. The hypotheses to be tested are Ho: 44 =0 versus Hype #0 Critical region is t < —tyo.025 = 2.78 and t > tyoors = 2.78 The test statistic is See aana7ys = ~ 16. Thus Ho is not rejected implyng that there is no significant difference between the two methods of analysis. Activity 6.10 1. A dietician told her clients that a certain diet was capable of reducing a person’s weight by as much as 10.0 kilograms, on the average, in a period of one month and provided that the person followed the prescribed dietary \. regime. To assess the validity of this claim, the weights of seven women who! followed the diet were recorded before and after a one-month period. The data obtained are displayed below. before 129 133 136 152 141 138 125 ofter 130 121 128 137 129 132 120 Assuming that the data are normaly distributed, test the dietician’s claim at the 5% significance level. 2. (a) A manufacturing plant discharges liquid waste into a river after treating it. An environmentalist claims that the treatment is not effective. In a study of the effectiveness of the treatment, 12 samples of water were collected up- stream from the discharge point, and 12 samples were collected downstream. aThe following data give bacteria counts and other descriptive statistics for the 24 samples. N MEAN STDEV SEMEAN downstream (Y) 12 35.2 6.9 1.99 2 ‘upstream (X) 12 298 68 1.84 ” Y-Xx 12 54 3.2 0.92 Assuming normality and equal population variances test the environmental- ist’s claim at the o = 0.05 significance level. () The company, which operates the plant, maintains that the treatment is ef- fective. In its study, bacteria counts were made from samples taken from ‘a number of sites downstream, prior to treatment. After the liguid waste had been treated, new samples were taken from the same sites and bacteria ‘counts again obtained. The results of the analysis are summarized below. N MEAN STDEV SEMEAN before treatment(Y) 10 37.6 6.4 2.02 after treatmen{X) 10 33.5 5.8 1.83 Y-X w 641 «12 0.34 Test the company’s claim at the a = 0.05 significance level. You may assume that the data are normally distributed. Comment on the results of the two tests in parts (a) and (b)? 6.11 Tests Concerning Two Proportions Finally, very often we would like to test some hypotheses about two or more proportions. In this section and in this course we shall learn how to tést for a difference between two proportions. the important thing to note here is that the test statistic used here takes a slightly different form from the one used in constructing confidence intervals. Let p; and pz and be two proportions of individuals or items having a certain char? ‘acteristic, in two populations. respectively. Suppose also that we would to test the hypotheses Ho: ‘Then the test Statistic is =p. versus Hy: not Ho Z= Pi-Pa via 9) [ze where i = ki/mi, Bi = ha/na, and p= Bete ay Example 6.14 The total number of female births and total number of births in a ‘certain country during 1988 and 1989 were recorded as follows year female births total births 1988 120 200 1989 240 500 92 LTest the hypothesis that the proportion of female births in 1988 was significantly higher than that for 1989. Solution 6.13 Let pea and pao be the proportions of female births for the two years respectively. Then the hypotheses to be tested are Ho: Pes = Pao against Hy : pss > pao ‘The sample estimates are _ 120-240 = 200+ 500 ea = 120/200 = 0.6, pao = 240/500 = 0.48, 6 = 360/700 = 0.51. The test statistic is Z 2.9. The critical region is Thus Ho is rejected and conclude that the proportion of female births in 1988 was significantly higher that that for 1989. 6.12 Summary of this Unit ‘This Unit brings us to the end of this course. In this Unit we have learnt the steps and procedures of statistical hypotheses i.e. making statistical decisions. The basic steps in Hypothesis are to formulate approraite hypohthesis, collect relevant data or if the data is already available, compute the test statistic. The general form of the statistic is estimate - true value std of estimate ‘The next steps are then to choose or decide on the level of confidence to use, determine the critical region and then make a statistical decision on whether or not to reject the null hypothesis Ho. A final conclusion concerning the parameter(s) of the data must then be made. T= Exercise 6.1 1. Let X1,...;X25 be a randon sample from N(y,25) and suppose that one wants to test the hypothesis Ho:"=0 against Hy : 4 > 0 Determine a suitable test statistic and ‘best’ critical region of size a = 0.05, for this test. If one finds @ = 1.6 determine whether or not, Ho should is rejected. 2. A telephone cable is being manufactured by two processes. It is desired to de- termine whether the process used affects the mean breaking strength of the ca- Ble. Suppose laboratory tests are performed by putting samples under tension and recording the load required to break the cable, and that the following data are obtained. sample sample size average load (kg) variance x 6 8.2 2.0 Y 7 11.2 4.0 Assuming normality and equal population variances test the hypothesis that there is no difference between the mean breaking strengths of wires produced by the two processes. 93 ll3. In an experiment to study the dependence of hypertension on smoking, a ran- dom sample of 180 individuals was selected and their smoking habits recorded as follows: smokers non-smokers hypertension 66 21 no hypertension 45 48 Test at the 5% significance level, the hypothesis that incidence of hypertension is higher in smokers than in non-smokers. Further Reading 1. Mood A.M, Graybill F.A and Boes D.C (1974), Introduction to the theory of Sta tics, Mc Graw Hill, New York. 2, Hog R.V. and Graig A.T. (1978), Introduction to Mathematical Statistics, Macmil- lan Publishing co. Inc, New York. 3. Goeneveld R. A. (1988) Introductory Statistical Methods, An Integrated Ap- proach using Minitab, PWS-KENT, Boston. 4. Walpole R.E and Myers R.H, (1972), Probability and Statistics for the Scientists and Engineers.Macmillan, London. 946.13 Solutions to Exercises i 1. Z= V25(X - 0)/5=1.6. Critical region=(1.645, 00). | ‘Thus Ho is not rejected. 2. T= (8.2~11.2)/ /(84/11) (16+ 1/7) = ~0.3.067. Critical region is \ (-c0, -2.201] U [2.201, 00). Ho is rejected. 3. Ho: ps =pw, Hy: ps > py. Z = 3.7894. Critical region=(U1.645, 00). Ho is rejected. Conclusion: Incidence of hepertension is higher in smokers. 95