CHAPTER 8 : ANALYSIS OF VARIAF{CE Sub-Topic * One Way ANOVA Chapter Learning Outcome + Solve the problems involve the analysis of variance. , n Learnins Obiective By the end of this chapter, students should be able to {' Use the one-way ANOVA technique to determine if there is a significant difference among three or more means. Develop an understanding of the underlying rationale by studying the example in this section. Become acquainted with the nature of the SS (sum of square) and MS (mean square) values and their role in determining the F test statistic. * Be aware of several different analysis of variance designs and understand when to use each one. Kev Term (Enelish to Bahasa Melavu) English Bahasa Melayu I Variairce --, Varians ) Treatment Rawatan J- Significant difference -t Perbezaan yang bererti ,] * {. 335 Chapter 8: Analysis of Variance Previously, we haye discussed methods of testing whether .,!. Two populations have equal means (independent two-sample test). * The means from two measurements on one populations are the same (paircd tests) ANOYA stands for AlYalysis Of YAriance, allows us to extend this to more than tm populations or measurements (treatrnents). That is, we can test the following: {. Are all the means from more than 2 populations equal? {' Are all the means from more than 2 treatrnents in one population equal? (This is equivalent to asking whether the treatments have any overall effect) To set our notation, let I be the number of populations or treatments being comparod and let7.t, ,p2wFi be the I means. Then the hypothesis for testing is Ho: pt - ltz : ...: pi H, : at least 2 of the p's are different To test these hypotheses, we are required a random sample from each population c treatment. The test used in ANOVA compares the variation (measured by tb variance) between the treatrnents (population) with the variation within trea&errr (populations). If the between variation is much larger than the within variation, fu means of difference treatments (populations) will not be equal. If the between wifth variations are approximately the same size, then there will be no significm difference between treatment (population) means. The z and , tests should not be used when three or more means are compareL The reasons why the e and ltest should not be done are: {' when one is comparing two means at time, the rest of the means under shdt are ignored. With the Ftest, all the means are compared simultaneously. 336 Chapter 8: Analysis of Variance when one is comparing two means at a time and making all pair-wise comparisons, the probability of rejecting the null hypothesis when it is true is increased, since the more f tests that are conducted, the greater is the likelihood of getting significant differences by chance alone. the more means there are to compare, the more t tests are needed. Even though one is comparing three or more means in this use of the Ftest, variances are used in the test instead of means- The ANOVA test is based on three assumptions: * The populations from which tlre samples were obtained must be normally or approximately normally distributed. {. The samples must be independent of one another. * The variances of the populations must be equal. 8.2 One Way ANOVA ,' Definition I One-way analysis of variance is a design in wtrich independent samples are obtained from * levels of a single factor for the pufpose of testing whether the k levels have equal means. We called it one-way analysis of variance because we use a singls property, or characteristics, for categorizing the populations. This characteristic is sometimes referred to as a treatment, orfaetar. Definition 2 The term treatwent is used because early applications of analysis of variance involved agricultural experiments in which different plots of farmland were treated with different fertilizer seed types, insecticides, and so on. * * 337 Chapter 8: Alalysis of Variance Definition 3 Afactor refers to a quantity under examination in an experiment as a possible cause of variation in the response variable- Delinition 4 Level is referring to the categories, measurements, or strata of a factor of interest in the current experiment. ! The method of analysis of variance us based on this fi.rndamental concept: With the assumption that the populations all have the same variance o' , we estimate the common value of o2 using two different approaches. The F test statistics is the ratio of those estimates, so that a significantly large F test statistics is evidence against equal population means. The two approaches for estimating the cofirmon value of o2 are as follows: t?' The variance between samples (also called variation due to treafinent) is an estimate of the common population variance ozthat is based on the variation among the sample means., {' The variance within samples (also called variation due to error) is an estimate of the corlmon population variance o2 based on the sample variances. Theor:v I Analysis-of-variance table : Source of variation Degrees of freedom Sum ofsquares Mean square f Treatments (between) k-1 SS(Ir) or SS(between) MS(fr) or MS(between) MSVtl MSE Error (within) N-k SSE or SS(within) MSE Total N-1 ssr 338 Chapter 8: Analysis of Variance SS (total) = SS (tre atment) + SS(error) SS(treatment) and SS(error) aie both sums of squares, and if we divide each by its corresponding number of degrees of freedom, we get meons squares. Some of the following expressions for mean squares include the notation N: N: total number of values in all samples combined MS(treatment) is a mean square for treatment, obtained as follows : MS(treatment)= SS(treatment) k-1 MS(error) is a mean square for error, obtained as follows : MS(error)= SS(error\ N -k MS(between)= SS(between) k-1 MS(within)= SS(within) N -k Ms(total) is a mean squffe for the total variation, obtained as follows : MS(total'1= SS(total) N-1 O _ vari?ncebetween samples vanancewithin samples Method I : Procedure Table for finding the Ftest value for the Analysis of Variance Step I : Find the mean and variance of each sample. (x,,r,' ),("r,r; ),-.,(x r, 11) Step 2 : Find the total mean. sr v /J-- Ar t : --'-- N Step 3 : Find the between-group variance. SSo"*""r:4Hd 339 Chapter 8: Analysis of Variance Step 4 : Find the within-group variance. SS.,/,i, = Step 5 : Find the F'test value. . - SS u"*,", SSrun,, The degrees offreedom are d.f.N: ft -1 where kis the number of groups d.f.D: N -k where N is the sum of the sample sizes of the groups, N : nr+ nz + ...* flt. Method 2 : Procedure Table for findins the F test value for the Analysis of Variance Stepl:FindSSr,,,t (>a' +ZB2 +2C21- (l'<+:r+:c)' \/n Step2:FindSSn"*"* l$ oY* (:r)t + (r cFr -(z e + z ! + zcY Step3:Find SSw*,in SS within : SS trtot - SS u"*"", Example 1 A researcher wishes to try three different techniques to lower the blood presswe of individuals diagnosed with high blood pressure. The subjects are randomly assigned to three groups; the frst goup takes medication, the second group exercises, and the third group follows a special diet. After four weeks, the reduction in each person's blood pressure is recorded. At alpha 0.05, test the claim that there is no difference among the means. The data are shown. 340 Chapter 8: Analysis of Variance x2 sl Jrr sf = 11.9 =5.7 = 3.8 =l$.2 7z =7.5 si =10.3 Answer ExamDIe I Step 1: State the null hypotheses and identifu the olaim Ht = 14 = pz= p, (claiwj Ho =al least one meatr is different ftom the others. Slep 2 : Fiad the critical value. Sinse& = 3andN = l5 d.f.N:t_l = 3 _l=2 d.f.D:N *k=15_3=12 , The critieal value is 3.89, obtained from the Statistics Table witha = 0.05. $tep 3 : Compute the test value, using the procedure outlined here. t Find the mean and variance of each sample (these values are shown below the data). + Find the grand rnean. The gmnd mean denotedby X6*, is the mean of all values in the samples. xza{ Z,X l0+12+9+...+4 116 a 4^ =.'.....,...'.,._:-_-.--=-= t.t) .rr 15 15 When samples arc el1ual in size, find XGM by summing the-F's and dividing by,t, whereft : the mmrber of groups. + Find the betrveen-group vriance, denoted by "3. 341 Medication Exercise Diet t0 6 5 t2 8 9 I 3 t2 15 0 I l3 2 4 Chapter g: Analysis of Variance z,(x -x )P MSB:ff _ s(tt.s - t.tt)2 + s(3.s * t.ts)2 + s(7.a _ t.tt)2 3-l - 160 '13 = 80.07 2 * Find the within-group variance, denoted byri,. _Z(" - t)r,l -jG:, _ (s - rXs.z)+ (s - rXro.z)+ (s _ rlro.:) 104.8 =-lZ =8.73 This formula'finds and overall variance by calculating a weighted average of the individual variances. It does not involve using differences of the means- * Find the F test value ,s,s" 90.07 l< =-------L =9.17 ss, 8.73 Step 4 : Make the decision. The decision is to reject the nuil hypothesis, since g.r7 > 3.gg. Step 5 : Summarize the results. There is enough evidence to reject the claim and conclude that a least orc mean is different from the others- ANOVA Table MS* Within (eror) 342 Chaphr 8: Analysis of Variance Example 2 A state employee wishes to see if there is a significaot difference in the number of employees at t}te interchanges of three slate toll roads. The data are shown. At a = 0-05, can it be concruded that there is a significant difference in the average number of employees at eaeh interchange ? AnswerEla4nlpl Step 1 : State the null hypotheses and identifu the claim Ho = l\ = Fz= frs Ho =ut least one mean is different from the others. (clairn) Step Find .sS,*,,= {rn' +Is? *5cr}- E41E{:Igf = (lz6l + toz * r+oi- El 12 t I*I I = 823.555 Find ssa"*.,,=ff{Xry +(IBF +EcFl* El:zr + rc} JI}J Air Hitam Interchange (F Batu Pahat Bypass (B) Step2:Makearable A AJ B B' C CZ 7 49 10 100 I 14 t96 I I 12 144 32 1024 I I l 53 1269 t2 102 l4 146 J =Lls* +rz, :356.222 Chapter 8: Analysis of Variance (st +rz_+AY 9 + r+'z] - Find SS *ahio = SS,rrol - SS. u"*"r, = 823,555 -356.222 = 467.333 AFIOVA Table Step 4 : Make the decision. Since 2.28675 < 3.68, the decision is not to reject the null hypothesis. Step 5 : Summarize the results. There is not enough evidence o support the claim that there is a difft among the means. Source ofvariation 178.111 823.555 344 Chapter 8: A.nalysis of Variance Analysis-of-variance table : Source of variation Degrees of freedom Sum ofsquares Mean square f Treatrnents (between) k-1 SS(Tr) or SS(berween) MS(fr) or MS(between) MS$t MSE Error (within) N-k SSE or SS(wfthin) MSE TotaI N-1 s.sr SS(roral) = SS (treatment) + SS (err or) N: total number of values in all samples combined MS(treatrnent) is a mean squlre for treatment, obtained as follows : | *, torr* ent) = tt'";T""'' l' MS(between)= SS(between\ k-l MS(error) is a mean square for error, obtained as follows : I ,t,r*,or):q# l"' MS(withinr=!X:? Ms(total) is a mean square for the total variation, obtained as follows - _ variance between samples variance within samples 356 Chapter 8: Analysis of Variance Method 1 : Procedure Table for lindins the -Ftest value for the Analysis of Yariance Step 1 : Find the mean and variance of each sample. (r-, ,,,'),(r,,,:)"--,(r- ,,; ) Step 2 Find the total mean. sr L-- N Find the between-group variance. SSb"*r"n: Step 4 Step 3 Step 5 Find the within-group variance- f (, - tl'i rrwi,h,n = jG_T Find the Ftest value. p, - SS un*,""n SS*,,0,, The degrees offreedom are where ft is the number of groups where N is the sum of the sample sizes of the groups, N:t\+nz+.**n* Method 2 : Procedure Table for findins the F test value for the Analvsis of Variance Stepl:FindSSron,r ( ze' +z Bz +>.c21- (r'e + : r + zcI \-)n Step2:Find SSB"*"", d.f.N: fr-1 d.f.D: N -k k-1 357 lrtroY* (:rI * (rc)et -(ze+z! +zcY Step3:FindSSwi,ni, SS *ithin : SS to,ot - SS br*r", Chapter 8: Analysis of Variance 3s8