Hypothesis Testing

Download as pdf
Download as pdf
You are on page 1of 14
CHAPTER THREE Hypotheses Testing 3.1 Introduction It is an established fact that daily activities are based on decisions taken by individual groups, cooperate bodies or state. These decisions are as a result of sample information. This sample information is in making statement or assumptions or guesses about the population, which may or may not be true. For instance, one can make the statement that female students are better than the male folk in English language. The average score of BAM 2 students in mathematics is 25. If it is found that the result obtained from a random sample differs from what is expected under the hypotheses, we would say that the observed difference is significant and the hypothesis may be rejected. On the other hand, if the observed result does not differ with the hypothesis, we conclude that the hypothesis is true or that the sample does not detect the real difference. The steps or procedures, which enable us to check whether any difference exists between observed sample values and expected results are called test of hypothesis. 3 3.2 Types of Hypothesis The two types of hypothesis are the null hypothesis and the alternative hypothesis. The Null hypothesis being tested and is denoted by Ho. For instance, Ho: The average age of Accountancy students is 22 years. Ho: Female students do better in English language than their men folk. The null hypothesis is formulated for the purpose of rejecting it. Alternative Hypothesis This is any other hypothesis, which contradicts the null hypothesis. Hi: Average age of Accountancy students is not 22 Hj: Female students do not do better than their men folk. 3.3. Type I and Type II Errors Decisions to accept or reject in hypothesis testing is usually based on the result of the sample. In taking these decisions, we commit two types of errors. These errors are referred to as type J and type II errors. 31 32 Statistics for Higher Educa, sis when it is try Type I error is committed when we reject the null hypothe © ang should have been accepted. th Type II error is committed when we accept the nul and should have been rejected. 7 The probability of committing a type I erm a is denoted by a. For instance, if «= 0.01, this imp aa fa correct decision had been taken or that one in every type | error. ypothesis when it is Not true called the level of significance that we are 99% confident that th cases, we shall be commiting 3.4 Types of Tests We have two types of tests: i, One tailed test and ii. Two tailed test. One Tailed Test A Test where the alternative hypothesis is expressed as “less than” (S) or “greater thax? (>) is called a one-tailed test or one-sided test. Its rejection region lies only on one sid: of the distribution. The level of significance a. is used as given in determining the critica t region. ‘Two Tailed Test . A test of any statistical hypothesis where the alternfatiye hypothesis is expressed as “nt equal to” (#) is called a two-tailed test or two-sided test.” It has two rejection regions ait each lies to the left and to the right extremes of the distribution. The level of significa ais divided by two whlle feverminine the critical regions. 3.5 The Critical and Acceptance Regions The critical region (rejection region) is a set of possibl sti leads to the rejection of the null hypothesis. ‘On ihe ther Reeder oy i a a set of possible values of the statistic that leads to.the acceptance of the are al The value of the test statistic that separates the critical region and the e null hyp: ical called the critical value. The value depends on the size of th acceptance reg sometimes on the degree of freedom of the test. le critical region Hypotheses Testing * Figure 3.1 Critical region Critical region ee q Hy: = W> Ho Hut < Ho Critical region Critical region , ih H# Ho 3.6 Test for One Population Mean ‘A one sampte'test can also be referred to as a test of hypothesis for one population mean. The purpose/of the test is to determine if the population mean is less than or greater than. or equal to some given value. This test could be done by taking the following steps into consideration: 1. When the samples are drawn from a normally distributed populations with known population standard deviation. 2. When the samples are drawn from normally distributed population with unknown standard deviation and the sample size is small. 3. _ When the samples are drawn from a population that is not normally distributed. Case I: Known Population Standard Deviation Suppose we want to test the null hypothesis that the mean p1 of a population with known standard deviation 6, is a specified values versus the appropriate alternative hypothesis, we shall draw a random sample of size n from the population and calculate the sample mean x. ? The test statistics is: EH, OG, Un sti High . 34 statistics for ? ligher Educatigy, The critical region depen, ts . i one. ~ is normally distri i an zero and variance ure of the altematj, ly distributed with me: paee depends on the nat : ‘Mative on the decisi : alternative being tested. The de ioe stated as follows: hypothesis and the test statistic. This cat Accept Ho or resume Alternative Reject Ho if judgement if Hypothesis, — Fa 1: > Ho é H< Ho “ H# Ho 3. i Example 3.1 Demography is 40 with a standarg The mean score of year 1 statistics students in sco deviation of 3. 25 students belonging to.this class was tested and the men ag was 50, Can you conclude that the average score of these students has changed? Use a. = 0.05. Solution Required to test: Ho: p= 40 Hy: w# 40 Level of significance 0 = 0.05 Test statistic: Mn Decision rule: Reject H, if 2, > fy otherwise do not reject Ho. l, ; Critical value: 9); = 1.96 Computation: Hypotheses Testing 6 Conclusion: Since =~ \o.0 + =1.96. we reject 1, and conclude that the average score of the students has changed. Case II: Unknown standard deviation n <30 If samples are drawn from a population with unknown variance, we estimate the variance x vn thas the t-distribution with n— 1 degrees of freedom. Example 3.2 A random sample of 60 female tested for high blood pressure give a mean of 80 and variance 100. At @ = 0.05 test, the hypothesis that the mean is less than the specified population mean of 200. Solution Required to test Ho: 4s = 200 Hy: w< 200 sz Q Level of significance a = 0.05 Test statistic: Critical value: coos = 1.65 Conclusion: Since /Z,/ =9.30 >, =1.65, we reject Hy and conclude that the mean is less than 200. 36 Statistics. for Higher Educatig, 3.7 Test For Two Population Means ne difference between the performancg A Most often, an investigator may wish to know # dents. He may want to determing the Science students in mathematics and the arts students. MY Tr test for significas Spending pattern of Aba traders and the civil servants 10 ull hypothesis specifying thay difference between two population means, We oe a could be stated as, there is no significant difference between the two. The hyp a Ho : pty = py (there is no significance di fference) b. Hy : [ty > pt (one mean is better than the other) en bd Hy : 11; # 2 (the mean of one population is not equal Case I . lying the sampli When the population standard deviations 0; and 2 are known and a Ee : ne 8 pling distribution for difference in means for two independent samp! 5 2, We compute the sample means ¥, and X,. The test statistic for the difference between two population means is given as: which has a normal distribution with mean zero and unit variance. Example 3.3 A standard taxation test was given to 45 male students and 65 female students. The male students made an average grade of 57 with a standard deviation of 4 while the female students made an average grade of 70 with a standard deviation of 6. Assuming that the students scores in the test are normally distributed, test the hypothesis that the male and female students perform equally well on the standard taxation test. Use a = 0.05. Solution Required to test: Ho: Hi — 2 = 0 Hy: wi -p2 #0 Level of significance a = 0.05 Test statistic: Uypotheses Testing 7 Vay n, Decision rule: Reject Ho if =, , otherwise, do not reject. Computation: 57-70 -13 _ B 0.9536. 3 Critical value: 2,7 = Zips = 1.96 Conclusion: Since | 2, | =-13.63, we reject Ho and conclude that the male and JSemale students did not perform equally well on the standardized taxation test. Type IT When the population standard deviation o; and o2 are unknown and the sample sizes nm and nz are small (i.e. n < 30), where the two samples are assumed to be normally distributed with equal standard deviations then the test statistic for testing the difference between two populations means is given as: which has a t-distribution with n; + nz —2 degrees of freedom. Example 3.4 In a survey, to evaluate the potency of an NPIC fertiliser on two variety of crops and determine if there is any significant difference between the yields of these crops, two independent samples were planted and the fertilizer applied on them. Information obtained from the yields of the two samples is stated below: Variety A Variety B na=15 ng =12 sa=5 sa=6 x, =81 X, = 64 xa -yalue. That is, Ho: 7 = %» and the possible alternatives e. 0 statistics for Higher Ec ducati, 38, ce usINS a = 0.09. «at aa ifferen Determine if the mean yiclds show any significant diffe Solution Required to test He: My — Hs =9 Hi: 4-H #9 Level of significance or = 0.05 Test statistic: Decision rule: Reject Ho if ¢, > otherwise, do not reject. Computation Y _ 81-64 =8,035 14xS +Hxe (1 1 15+12-2 Us: 12 Critical value: 3%,, = 2.06 7D 0 2.06 Conclusion: Since te = 8.035 > tooas = 2.06, ; : significance difference. s 2.06, we reject Ho and conclude that there is * 3.8 Test for One Population Proportion : In binomial experiments or other experi Sie F . ‘periments ‘where ‘ ; outcomes, we require testing the hypothesis using me exis only two possible testing the proportion of successes in the binomial oa . Interest may be usually it eriment equals if uld be: ome spec @ Hypotheses Testing Hi: w> ny Hi n Zu otherwise -do not reject. Fo statistics for Higher Edy cay Computation p= =2-057 -1.96 0 2.06 1.96, we reject H, and therefore conclude that the Conclusion: Since /2,/ =3.46 <2, student’s claim is not valid. Example 3.6 Some patients claimed that a malaria drug taken by them causes at least 20 percent of cough and catarth. If 50 patients that were randomly selected claimed that 12 of th had cough and catarrh, can the null hypothesis be rejected at.@ = 0.01? / Solution | Required to test: ‘ Hy: #2 0.20 Hy: 2 <0.20 Level of significance a = 0.05 Test statistic: zy Hypotheses Testing Decision rule: Reject Ho if 2, > 2, , otherwise do not reject. Computation: 2.33 Conclusion: Since 2, = 0.67 <2, = 2.33, we do not reject Ho, we therefore conclude that the malaria drug causes at least 20 percent of the cough and catarth. 3.9 Test for Two Population Proportions Certain situations arise when we may wish to test if the two population proportions 7 and 7 are equal. Here x; and x are the number of successes in population 1 and 2 with sample sizes n; and n3 respectively. The unbiased estimates for the population proportions are: x, py =Land py == ny ny We may wish to test the null hypothesis that 7) - ™ = 0 or m = 7 against any of the following alternative hypothesis. Hy: m - #0 Hy: tm - 12,<0 Him -m>0 when Ho is true, then the test statistic becomes: Sa On 42 tion of successes js computed as; The pooled estimate for the two population propo ae My Py +P: Ny nyt of successes in the first and second samples. Th nis: where x; and x) are the number 5 test statistics for equal population proportio _ = P:)--*) PiA= Pi) , Pall 7, n; which has a unit normal distribution. Example 3.7 ‘A sample of 150 fresh students and 100 final year students of Abia State Polytechnic Abs were asked to state their opinion whether they favour scrapping of JAMB that permits cach institution to admits its own students to alternative admission problem. 90 of th favour of scrapping the JAMB, fresh students and 66 of the final year students were in Use the 5% level of significance to test whether or not the two groups of students differ in their opinions on the scrapping of the JAMB. Solution Required to test: Ho: 71 = 72 Hy: 1 #7 Level of significance o = 0.05 Test statistic: (r= p2)- Ps P2ll= ps) n, n, Decision rule: Reject Ho if 2, > ®y, » otherwise do not reject. Hypotheses Testing ad Computation: Z No Critical value: -196 0 1.96 Conelusion: Since /z, /=0.94< yo =1.96, we do not reject Ho. We therefore conclude that there is no ‘significant difference in the proportion of fresh and final year students who favour the scrapping of JAMB. Exercise Three 165 A sample of 144 smokers is selected and 36 are found to prefer Benson and Hedges cigarette. Use 90% confidence interval to estimate the fraction of smokers who prefer Benson and Hedges. 2. 8 pairs of brothers scored the following marks in examination: Ojo 86, 81, 67, 71, 50, 58, 77, 66 Ikashi 79, 81, 58, 65, 91, 45, 82, 59 Do these data indicate that Ojo did better than Ikashi use a. = 0.05. In a sample of 200 seeds, 96 germinated. Would you agree that 90% or less germinated at1% level of significance? 42 statistics for Higher Educa, er from a parti Suppose we suspect that students with blood group Ondents were tala disease than those with blood group A. And SUPP rest to blood group A, mn blood group. 250 belonged to blood group O and the FE 199 studenis students of blood group O suffer a certain type of se 0.05, are we justified blood group A suffer from the same disease. Using & = -L> bo uphold the researchers suspicion? - i follows: The body weight of a sample of 5 men before and after a diet are as 1ol’ows: Before 76.1, 75.3, 75.4, 80.6, 83.7 After 80.2, 78.4, 74.5, 76.2, 81.2 am Do these data provide sufficient evidence to indicate that the diet was able tp reduce body weight? Use a = 0.01. i ‘

You might also like