8; Non-Parametric Tests INTRODUCTION Sampling tests, discussed so far, are known as parametric tests because these are based on the assumptions that the concerned sample has been obtained from a population with known values of its one or more parameters. For example, the use of t-test to test the Hy : 4 =H requires that the two samples are drawn from normal population with equal variance (o3 =03).Similarly, the use of Fiest requires to test Hy :0j =o} assumes that various samples are obtained from normal population with equal variance etc. The validity of the results of a parametric test depends upon the appropriateness of these assumptions. Thus, when these assumptions are not met, the parametric tests are no longer applicable. In such cases, it is essential to study non-Parametric tests. MEANING OF NON-PARAMETRIC TESTS. Non-parametric tests do not require any assumptions about the parameters or about the nature of population. By non-parametric tests we mean those statistial tests which dq not depend either upon the shape of the distribution or upon the parameters of the population mean, standard deviation, variance, etc. The assumption as to the normality or symmetricity of the population distribution from which the samples have been drawn is not required for these non-parametric tests. Non-parametric tests are sometimes referred to as distribution free tests. In addition to this, these non-parametric test do not require measurements so strong as that requires by parametric Tests. DIFFERENCE BETWEEN PARAMETRIC AND NON-PARAMETRIC TESTS Q) Inpar. i mptions of normal population is taken whereas in non-parametric test Fo ah aio is taken about the population. It is because of this that non-parametric tests are known as “distribution free tests”. B @ Parametric tests comprise of t-test, Z-test and F-test whereas non-parametric, tests comprise Chi-square test, Sign test, Median test, Wilcoxon signed rank test, etc. - ®) In case of 2 parametric test a normal hy is is set up and the variable is tested for drawing inferedtes-Tircase oF non-parametric tests, inverse (opposite hypothesis) is set up. DVANTAGES OR USES OF NON-PARAMETRIC TESTS ee ws a ‘e advantages (or uses) of non-parametric tests are mentioned ie any assumption 10 ne ‘on-parametri istribution free ie., they do not requ “tn tric tests are distribution free i “swathes distribution. ide about ion following normal or i 2) Carey or uation FoTowing none popes meena Genally, they are simple to understand and easy to apply ci aNon-Parametrie ctests donot require lengthy and laboroius arithmetical computa, (3) Most non-parametti i nsuming. ; 2 ana fewer and less restative assumptions than do the (4) ‘Non-parametric tests make fe rn aes seno alternative © using @ non-parametric test if the data are available in ording (6) There isn te nominal scale- ts are useful to handle data made up of samples from, - etal -parametric tests aft : ° Mpulations without making assumptions. TYPES OF NON-PARAMETRIC TESTS ' There are many types of non-parametric tests. The important among them are : (a) Chi-square Test (discussed earlier in 72 Chapter) (2) Sign Test 3) Wilcoxon Signed Rank Test (4) Mann-whitney U-Test (5) Kruskal-Wallis Test (6) Wald-Wolfowitz Runs Test : (7) Other Non-parametric Tests (Rank Correlation, Median and Kolmogorov-Smimov tests) Non-Parametric Tests Chisquare| | Sign Wilcoxon Mann- Kruskal- Wald- Otter Test Test Signed Whitney Wallis Test | | Wolfowitz | | Tests Rank Test U-Test Runs Test Let us discuss them in detail (1) Sign Test : The sign test is the simplest type of all the non-parametric tests. Its name comes from the fact that it is based on the direction or the plus or minus signs of observations ina sample and not on their numerical magnitudes. ‘Types of Sign Test The sign test can be of two types : (a) One-sample sign test, and (b) Paired-sample sign test (a) One-sample sign test : In one-sample sign test, we set up the null hypothesis that +and- signs are the values of a random variables having the binomial distribution with P= “* ol Hy :p=-or that = 1g Procedure : Ths test involves the followi Proce n wing steps : (Find the + and— sign fr the given distribution. Put a plus (+) sign for 2 valve themean value(u9), a minus (~) sign fora value smaller than the mean valueand 922 os value equal fo the mean value. ii) Denote the Signs (j em he tal number of signs (ignoring zeros) by n and the number (iii) Obtain the critical value (K) of les: by using the following formula : of les eave of signifi frequent signs (S) preferably at 5% level_ ee ic Tests oneParametn® 229 0:98 vn the value of ‘S’ with the critical value (K). If the value of S is greater than the jv) compare then : Ge) Vaiue of K (Ee, $>®) the null hypothesis is accepted. If $< K, the mull hypothesis is pag ee perfidy ig Be lity Dis . When sf all (ie.,2<25), we find probability of the rete signs p(5) by the sum of the probability of S of fewer $ using the binomial distribution -* »* withp=>- a nc p* with P=> Then, we compare the above calculated value of probability with fed value at 5% level of significance i., at a=0-05 for one tailed /two tailed tests. If the on a i aa : the ot probability P(S) is <0-05, null hypothesis is rejected and if P(S)>0-05, the null eee is accepted. e1. The production manager of a large undertaking randomly paid 10 visits to the bam worksite in a month. The number of workers who reported late for duty was found to be 2, 4, 5, 1, 6, 3, 2, 1, 7 and 8 respectively. Using the sign test verify the claim of the production supervisor that on an average, not more than 3 workers report late for duty. Use 5% level of significance. sist: ‘Let Hp : HS Sagainst Hy :u> Sat 5% level of significance. Determination of Signs w.r.t. x Signs (X -3) 2 - i : No. of Plus Signs = 5 5 ] No. of Minus Signs = 4 1 = 6 + No. of Zero =1 2 : Total No. of Signs = " = 9 2 - (ignoring 0's) a 7 . 4 7 + 8 + From the above table, we get Total no. of signs (ignoring zeros) = " Number of less frequent signs = S = 4 aoe : The critical value (K) of less frequent signs (5) at 5% level is given by : ry K BA 99g fie 5 0:98 V9 =4-2-98=1 08 a the critical value of K(1-06) the sample data support the 39 Since number of less frequent signs $(4) is more on ie, $> K, the null hypothesis is accepted. It means claim of the production supervisor. i ility Distribution. Aliter: We a solve the above problem using Binomial Probability tienExample 2. Solution: 1 20.05, 5=4 Weare given, a9, Paya Oe ersuccess is given by : The probability of or fewer st Be ia 1p (2)*4%0(3) (2) #7eaC3) (3) Piss4]= °C, G a a(S) Go vei(a) (2) +°eo(3) (3 9 19 /1)9 steal a) OG win af 6) 9G 1 256 195 ay tase 08+ 364945192 itis =-05) From the above, it is found that P(9>ala This suggests that the null hypothesis is accepted. It means that the sample data support the claim of the production manager. laying four rounds of golf at the city club 11 professionals totalled Sa, 250, 273,283, "283, 275, 284, 282, 279 and 281. Use the sign test at 5% level of significance to test the null hypothesis that professional golfers average 11=284 for four rounds against the alternative » < 264 =284 against H :1-<284 at 5% level of significance. Let Hy Determination of signs wt. w=284 Tx: 280 | 282 273 | 283 | 283 | 275 | 284 | 282 | 279 | 281 sions (x-29 | - | -|+]-[-|-]-jo]-/-|- From the above table, we get No. of Plus signs = 1; No. of Minus signs = 9, No. of Zero = 1 Total no. of signs (ignoring zero) = n =10 Number of less frequent signs (#) = $= The cial value (X) of less frequent signs (5) at $% level is given by n- 10-1 = 70-98 Vi=—— 0-98-10 = 45 - 3-099=1-40 Since Sis less than critical value of K (1-40) i, S K, the null hypothesis is accepted. This mea" that there is no signi eee eee the new collection policy in accounts receivable. difference before and after # ‘NIMPORTANT TYPICAL EXAMPLE . lone continuously for 7 cular exercise if d 4, A physical instructor claims that a partis se if lie reduced weight by3-5kg. Five overweight gitls did the exercise for 7 days and their weights were ‘observed as under: : : Girls : 1 2 x n a Weight before exercise = 70 2 2 . 66 70 7m ise reduces Weight after exercise : Making use of the sign test, verify the claim at the weight by at least 3-5 kg- a =0-05 that the exerc!Non-Parametric Tests weight before and after exercise by }1y and H respectively, we olution: Denoting the mean have the following, Hg thi ~H2 2 3-S against Hy : Hy ~H2 <3°5-4= 0°05 Determination of Signs w-r.t. 3.5 Girls x x D=(x -¥)_| Signs (D-3.5 1 7 66 4 a 2 m2 70 2 . 3 75 n 3 . 4 n 6 5 : 5. 78 2 6 a From the above table, we get No. of Plus signs = 3; No. of Minus signs = 2 ‘The total no. of signs orn=5 Number of less frequent signs = ritical value (K) of less frequent signs (5) is given by: $=2The cl 2g inp 09852-21970 Since $>K, Ho is accepted. It means that the sample data support the claim that Fine ercse if continuously done for 7 days reduces the weight by at least 3-5kg. LARGE SAMPLE AND SIGN TEST When the sample size is fairly large (ie, n>25), we ust the normal approximation to the binomial distribution to carry out the sign test. The value of ‘2’ can be computed as: S-np 2 Inp (1-p) ‘Then we get the critical value of Z at the desired level of significance. If the caleulated value of Z happens to be less than the critical value, then we accept the null hypothesis. Ifthe case is reverse, then we reject the null hypothesis. Given below are the data relating to the daily milking from a cow for 30 days: Example 5. 2% | 20-8 |18-6| 16-6 | 23-2] 21 | 19-2 | 24-8] 23-8 | 2-6 Milk in Litres | 19 | 23-2 | 20-4 | 22-8 | 24-6 | 22-8 | 20-8 | 22-6 | 21-4 18-6 18.6 | 17-4 | 19-2 | 20-4 | 22-2| 24-6| 20 | 22-4] 20 | 3 Using sign test to test the null hy i i ypothesis at 5% level of significance that the average daily milking from th 4 Hi i ternati ee a Solution: We have, Hg :1=22-4, Hy :1<22-4, a=0-05 (left tailed test) Determination of the signs w.r.t. 22-4ove, we get "No. of Minus signs = 18, No. of Zeros = 1 igns ignoring 2er0) ‘The no. of less frequent signs (+) ize of samples is quite large (n i.e., n>25), the test which is a close since the si 7 - approximation 10 the Binomial distribution is used. Ce a hte inp =P) Substituting the value in the above, we get n-2(5 1-145 ‘The critical value of Z at 5% level for left tailed test =-1-645 since the calculated value of Z Rt or ER” ranks which given at the end of the book then reject the nulllees Non-Parametric Tests = i large undertaking randomly paid 10 visits to the - lial ster aeee ri beret workers reported late for duty was found tg Morksite ina month i and 8 respectively. Using Wilcoxon's signed-rank test be £2 4 5 6.» the production supervisor that on an average, not more than 3 Sarde Teport late for duty. Use 5% level of significance. : ae Hy:u>3 Solution: Let us take the hypothesis that Hp : = 3.against Mi: h> Seagrers al Ranks (+ ignored) Signed Ranks | R* R 5 = Fi 2 = 2 4 z 1 . : = 5 2 2 a . 17 | -2 2 2 5 5 oe é : 7 7 e 3 o 0 = = 2 -1 1 : > 2 1 -2 2 2 s s : : i 8 8 . ; 5 5 9 9 : Total 7 n=9 =R* =31 | ER =14 From the above table, it must be seen that total number of ranks =n=9(i.e.,<25), Since n<25, the test statistic is given by = ; T =smaller of the two sums of the signed-ranks = 14 Looking at Wilcoxon's T table at 5% level for one tailed test at n=9, we get the critical value of T or Togs =8. Since the calculated value of T (14) is greater than its critical value (8), the null hypothesis is accepted. It means that the sample data support the claim made by the production manager. : (ii) Two Sample Signed Rank Test (or Paired-Sample Signed Rank Test) : The Wilcoxon's signed rank test has important applications in problem involving paired data. Such a test is widely used by the research scholars in their study of two related samples or matches pairs of ordinary data viz., outputs of two similar machines, responses gathered before and after a treatment, etc. where we can find both the direction and magnitude of difference between the matched values. In these problems, we find the difference between each pair of values with algebraic signs. Then we proceed in the same manner as in the case of one sample signed rank test. Example 7. _ Use Wilcoxon's signed-rank test to see if there is a difference between the number of days until the collection of an account receivable before and after a new collection policy. Use the 0-05 level of significance. Before (X) :| 30 | 28 | 34 | 35 | 40 | 42 [35 | 38 | 34 | a5 2s | 27 | 25 | «1 | 36 After (Yd: | 32 | 29 | 33 | 32 | 37 | 43 | 40 [a1 [37 | a4 | 27 | 33 38 | 36 |f won wo ic Tests 235 Let us take hypothesis Hp : There is no differ i a before and after a new collection policy in the ae - number of days And Hy: There isa difference between the two ivable. Determination of Signed Ranks x Y |d=x-y| Jay Ranks (R) | R : rn een I Signed Ranks 30 32 -2 2 ; R R 2B 29 -1 1 3 a 6 a 33 1 1 . ; 3 35 32 3 3 7 : . 40 7 3 3 : : [ie 2 43 -1 1 3 : 33 40 -7 : 7 uu = 1“ 38 41 3 3 ; . : 7 = 4 3% 3 3 9 = 9 45 44 1 1 3 ; iq 28 7 1 1 3 : : a 33 -6 6 1B = . 25 30 -5 5 2 7 41 38 3 3 9 9 . 36 36 = = : = . Total n=14 ER* =36 | ZR" = 69 From the above table, it must be seen that the total number of ranks=n=14 (n<25). Since n<25, the test statistic is given by T =smaller of two sums of the signed ranked = 36 Looking at Wilcoxon's T table at 5% level for a two tailed test at n=14, we get the critical value of T =21 Since the calculated value of T is greater than its critical value, null hypothesis is accepted. It means that there is no significance difference between the number of days before and after a new collection policy. IMPORTANT TYPICAL EXAMPLE lar exercise if done continuously for 7 A physical instructor claims that a particul ‘girls did the exercise for 7 days days reduces weight by 15 kg. Five overweight and their weights were observed as under: Girls: = 2 3 4 5. Weight before exercise :_| __70 n 7s a 38 Weight after exercise : 6 7 2 =236 Solution: Non-Paramettic Tests Making use of Wilcoxon's Signed-rank test veri exercise reduces weight by at least 3-5 kg. Denoting the mean weight before and after exercise by 1 and 1 Tespectively, We have the following Ho ‘Hy ~ Hy = 3-S against Hy : 4-H <3-5,0=0-05 ify the claim at c=0-05 that the Girls} X Y | D=X-Y |d=D-3-5| |d| | Ranks (R) Signed ( ignored)| p+ pa 1 | 7 | 6 4 05 fos] 15 15 z 2 | 7 | 7 2 “15 }15| 3.5 . 35 3 | 5 | 2 3 -05 |o5| 15 . 15 a} | 6 5 150 115] 35 35 . 5 || 2 6 25 [25 5 5 z n=5__|=R*=10| er-=5 From the above table, it must be seen that the total number of ranks=5 (ie.,<25) Sincen<25, the test statistic is given by : T=smaller of the two sums of the signed ranks = 5 Looking at Wilcoxon's table at 5% level of significance for a one tailed test atn=5 we get the critical value of T or Ty 5 =1. Since the calculated value of T(6) is greater than its critical value (1), the null hypothesis is accepted. It means that the sample data supports the claim that the exercise if continuously done for 7 days reduces weight by at least 3:5kg, EXERCISE - 1 1, A teacher claims that by imparting coaching for one month we can make the student worth securing at least 50 marks. A random sample of 10 students reveals the following scores in the examination. Test the hypothesis that the claim is acce} ted. L_Marks | 52 | s4¢ | 60 | 65 Jas [a2 Tse | os | oo | a re re s | 9 | wo} [Ans. K=1.06, Hypothesis is accepted] 2. Anew car model was put to test course for observing kilometer run per litre of petrol. The data record were : 20, 18, 22, 21, 15, 17, 14, 16, 22, 23 Use sign test and Wilcoxon signed rank test to test company claim that the new car model on an average runs, 20 km per litre modal. eat : |; T =12-5 Hypothesis is accepted] [Ans. K=1-06 Hypothesis accepted; T=12-5 ope of days unt te ion of an account receivable before and after a new collection policy. Use at 5% 3. Use the sign test to see if there isa difference between the nu: collectic of significance. After PAAR 72176 | 60 | 66 | 54 | 33 | 74 | g2 | 80 (Ans. K=2-63, Hypothesis is acer |ie ae Non-Parametric Tests 237 4 eee are the number of patients treated by two doctors in a hospital during a By DociorX 14 18 {12 Tio [rs Tio [io [16 [a2 [8 ]o Jo Jn [is 17 ByDostory 118 lio |s_[9 Ta hz Ts fa Is lio le I7 la In Jo Use the sign test at 1% level of si ignificance, test the null hypothesis that on an average both ooo erore patenttaal Tumbers of patients as against alternative hypothesis ther be X treats more patients than Doctor Y. : 1,5,3,2,4,6,1 Using Wilcoxon signed rank test to verify the claim of the labour that an average not more than 4 workers met the accident. Use a=005, 6 Nine adults agreed to test the efficiency of a new diet program. Their weights (Ibs) were measured before and after the program - and are given below : Sr.No.} 1 2 a 4 5 6 |=% 8 9 Before | 132 139 126 ua | 122 | 132 142 19 126 After | 134 141 118 116 u4_| 132 145, 123, 121 ‘Use Wilcoxon match paired signed-rank test to test the efficiency of the diet. [Ans. T=15] 3) Mann-Whitney U-Test : This test was developed by Mann and Whitney and it is a rank sum test. This test is used to test whether two independent samples have come from the same (identical) population. ture : This test involves the following steps : () Arrange the data of both the samples in one column in ascending order. (@) Assign ranks to them in increasing (from low to high) order of magnitude. In case of the i) eaated Values, assign ranks to them by averaging their rank positions. - Then the ranks of the different samples are separated and summed up as R, and Ry. ) os om and " are sufficiently large (ie.,> 8) then find the test statistic U by the following Ny (ny +71) or My (My +1) Uy =nyny ++ —__R, i 2 = Myla 2 2 comparsa’® cttical value of U from the U-table, if available, with reference n, and n, for 4) if eo with the aby sient tbl ove calculated value. 5a 2 © is not available, then get the transformed form of the U-statistic which is z= —_U=(1y +11) /2 iy my (my +g #1) 7 1Dma 238 Non, vi) Then find the critical value of Z at 5% or 1% level of significance. : at 96 and at 1% is 2:58. ‘The critical vale ot vi) If the calculated value of Z is less than or equal to its critical val : eypothess or reject the same ifthe result appears to the reverse,” "°" ®0ept the ny Note:The value of U that we use for the Utes isthe smaller of Uy and Up Given below are the relating to production of rice in. [| Geax wenn uintals pet acre collecteg [Senet a [x6 20 [ae |26 [as Jae [20 Joa Jos Jos Jin Tg ampleB_|22 120 118 |22 |24 |32 116 log [ag Using the Mann-Whitney U-Test (a Rank Sum Test) at 5% i : that bh the samples have come from he poputons wi te Solution: Determination of the Ranks and their sums Values of both the samples in] Rank in Increasing [Ranks of the Ranks of ta ascending order order Sample "| SanpieB ts 8B 1 - 4 12-8 2 . -2 14-B 3 - 3 16-A 4 5 5 . 16-A 5 5 5 - 16-B 6 5 - 5 18-A 7 85 85 - 1&-A 8 85 85 - 18-B 9 85 = 85 18-B 10 85 = 85 20-4 ul 2 12 - 20-A 12 2 2 . 20-B 13 R n 2A 4 15-5 155 . 2B 15 15-5 2 22-B 16 15-5 a 2-B 7 15-5 . 2A 18 185 18.5 2B 19 } 18-5 - 26-A 20 20-5 205 26-A 2 205 205 28-A 2 2.5 25 28-A 23 25 25 Total 276 R,=171 N=3B m=———— sa Non-Parametric Tests 239 We have : Riv R= NW+D = 232 1714105 = 3 O3+0 176 verified Let us set up the Null : Population >. ‘ull Hypothesis, 1 : Both the samples have come from the the Almith the same mearen Mo : ternati, ‘ Population, -"“'¥e Hypothesis, 1, : The two samples are not from the same Since, both n 18nd" are > 8 the relevant test statistic U is given by Uy =mn, + aero -R, sxy 202+) *132+78~171 = 210-171 = 39 or U. 1 Myon, +2 +1) -R, 11 (11 +1) ee By transformation Of Uinto Zwe have, U=(n,. - = (y.m)/2___ 39 ie My 30, we compute test statistic Z based on Normal distribution by usin ()R=1- (when ranks are not repeated) (b)R=1- (when ranks are repeated) m—1 devel on et the ettial ical value of Z from the Normal Curve Table with reference 0 significane®metric TESTS el alculated valut of Z happens to be less than its critical value, then we accept the null 11th herwise we reject it. serving example, ilustrate the procedure of rank credit les. qe fl Use the rank correlation test at 1% significance level, determine if there is any poole 1 positive correlation between the prices of shares and prices of debentures given as below: Price of Shares (Rs) __|52_|53 a2 Too [as [31 [37 |38 [25 27 Price of Debentures (Rs) |65_|68 [43 [as [77 [4s [35 [a0 [25 [50 S Computation of Rank Correlation Coefficient se Price of R Price of Ry D=R,-R D2 Shares (X) Debentures (Y) 52 3 6 3 0 0 53 2 68 2 0 0 2 5 8B 6 -1 1 0 1 38 7 -6 36 45 4 7 1 3 2 31 8 48 5 3 9 37 7 35 8 -1 1 38 6 30 9 -3 9 3 10 3 10 0 0 2 9 50 4 5 25 n=10 =D=0 =D? =90 Applying the formula, rei -® =D? nen 6x90 540 21-3 = 1-5 =1-0-545=0-455, 10°-10 990 -01, R=0-455, ie, there is no correlation between the prices of shares and debentures) Stcey cap H'P>0 (there is positive correlation between the two) ang nie, 10) the critical value of R as obtained from the Rank Correlation Table for °=0-01 is 0.7818, Senut pe calculated value of R (0-455) is less than its critical value (0-7818), this suggests that een oottesis is accepted. Hence we conclude that there is no significant positive correlation Meg nce of the shares and price of debentures. Mai hea ‘an Test : It is another important non-parametric test. It is used to test whether two or PS are taken from the population with same median.~ 248 Non-Parametric Tesig Procedure : This test involves the following steps : (i) Set up the null hypothesis Hy that there is no difference in the median of the two sai (ii) We find the median of the combined data. Both groups are combined and analysed in ascending order and find median (M). (ii) We determine how many of the values in each sample fall above or below the median ( the two samples are classified in two groups above median and below median), Th frequencies are counted for each group. : (iv) We present the data in the form of a (2x 2) contingency table shown below : imples, data are Sample 1 Sample IL Total Above Median. a b a+b Below Median c d ced Total ate bed (v) We now calculate the expected frequencies by using the formula : (Row x Column) /Grand Total (vi) We compute the value of x? by using the formula : To-p?1 De Becta eee (vii) Determine the degree of freedom =v=(r-1) (c-1). (viii) The critical value of x? at 0-05 level of significance for given degree of freedom is found. (ix) If the calculated avlue of x? exceeds the critical value of x2, Hg is rejected. It implies that there is no evidence to suggest that the median is the same in case of two samples. In the reverse case, Hg is accepted. Example 2. Two different fertilisers were used on a sample of eight plots : PlotNo. 1 | 2 3 4 peeled oye ee Fertiliser 1 49 | @ | a | a xa | 3 | @ Fertiliser I 40 45 50 43 37 Az 55_| 97 Use Median test to test the hypothesis that the two fertilisers have the same ‘median. Solution: Ho : The two fertilisers have the same median. Now, we calculate the combined median of the two series. Let us arrange the data of two series in an asceding order : ir. No. Sr. No. 9 10 nu 12 13 4 15 16. fewanewne FSSSSRRSk tet249 n+1)% Combined Median = Size of( =) item : se =8-5th item 45 _/ 8th + 9th item n= 444 _ yy 5 2 2 Form a2 x2 contingency table: Sample | Sample IL Total | bove Median 3 5 8 [Blo Maan 5 3 8 Total 8 8 16 Now, calculate the expected frequencies : 8x8 EQ=— =4 E@)=8-4=4 E()=8-4=4 E(3)=8-4=4 ° E (0-5 (0-9? (0-5? /E 3 4 -1 1 0-25 5 4 +1 1 0-25 5 4 +1 1 0.25 3 4 =1 1 0:25 2 =E(O-B? /E=1 2 _»|O-b? x -+[Co9* Degree of freedom =v = (2 -1)(2-1)=1 The critical value of x2 at 5% for df. 3-84, Since, the calculated value of x? is less than the critical vale of x? at 5% 6 accept null hypothesis and conclude that the two fertilisers have the same Kolmogorov Smirnov Test Single ™°80FOV ~ Smimov test, named after statisticians A.N. Kolmogorov and N.V. Smimov, isa ae ed P*AMetric test for testing whether there is a significant difference between an ie cy ey distribution Values on Mare test of goodne aN ordi Proceduss ial scale. 1) sera Tis test involves the follow Los., we ‘median. ing steps : mead yah PONS tha here is no significant difference between the observed and and expert Vales (or theoretical values) ie,, there is good compatibility between theory ® on *Periment, Sis Of the null hypothes is, we calculate the expected frequencies.3) Compute the observed relative cumulative frequ : Non-Parametrc ro (9) Simutative frequencies (Fe). 7 lency (Fy) and (4) Determine the largest absolute deviations between Fy and R can Pected (5) Compute the critical value of D with reference to the val Fh Telative 5) Compyerov -smimnov Test Table given at the end of book. Mes Of m and & (Los) 7 core the calculated value of D with the critical value of D1 (6) Come to be less than the critical value, the accept Hy Inthe revere case ater : . a iste Example3, The following grades were given to a class of 100 students, ejected, Grade _|A c 7 ao Frequency 50 160. 20 40 30 f ‘Test the hypothesis that the distribution of grade is uniform. Use Kolmg Smirnov test. se Kolmogorov Solution. Let us take the hypothesis’ that the distribution of grades are uniform ie, there is no difference in their distribution. No. of students given grade = 200. We should expect = 40 to each student. © | Observed | Observed | E Expected [= Cumulative | Relative Cauaiitive| Ramive [>7F¥-%l Frequency | Frequency Fy Frequency |Frequency F, | 50 50 0-25 40 40 0-20 0.05 60 110 0-55 40 80 0.40 0.45 20 130 0-65 40 | 120 0-60 0.05 40 1m | (0-85 40 | 160 0-80 0.05 L30 200 1-00 40 | 200 1.00 0.00 From the table, we find that the largest absolute difference is 0-15 which is known as Kolmogorov - Smirnov D value. Since the sample size is more than 35, the critical value of D with referencetovalte of nanda is aad =0-096, As the calculated value of D exceeds the critical value of 0-096, we reject Hy and conclude that the grades are not uniformly distributed. EXERCISE-3 ———— : itive 1. Use the rank correlation test at 1% level of significance, determine i there an) PO" correlation between study time and scores : 2 Number ofhoursstudiedog:| 8 | 5 [||] s 16 +8765) Score(Y): ss [a Ln | 21m ee : jenificant relat [Ans: R = 0-9758, | Z|=2-94, Reject Ho and signetale tudes : 2. An LQ. test was given to a random sample of 15 male university. Their scores were recorded as follow: Male: _{56 166 62 {a1 |7s 2 fa fos fas fo LFemale: |63 177 165 |71 |74 |60 |76 161 167 a251 sei TOS! determine whether L.Q. of male and female students is same in the he median of the combined sample = 68) [Ans. x? = 0-252, Accept Hy] _ se table of observed frequencies, along with the frequency tothe observed under th s, Below ctribution : fate the K-S statistic. test to median et Us sil): (Given # Calcul Ve () Gare conclude that the distribution does in fact follow a normal distribution? Use 0) Cans and Kolmogorov ~ Smimov test ret 51-60 __ 61-70 71-80 81-90 91-100 quency 30. 100 440 500 130 390 100 = 0-117, Accept Ho] In what ways are they different from parametric tests. 1 Whatare non-parametric tests 7 and non-parametric tests, thus, highlight the advantages 1 Differentiate between parametric ofnon-parametric tests. 4. Discuss the methods of using ordinary sign test and Wilcoxon’s signed-rank test. 4. Bpplain Wilcoxon's signed rank test procedure. 5 ret various non-parametric tests. Describe by taking a suitable example Mann-Whitney Test. 6 Blain briefly the Chief features of Wald-Wo analysis. : Mite short note on Kruskal-Wallis H-test. t Whats sign test ? What is it used and what are its limit Eoplain the procedure of : q One sample sign test i oe signed-rank test SMa What olfwitz run test estate non-parametric tests ? Briefly describe the process of Wilcoxon's Signed-Rank "Wie shor notes on: Rank : Oi) gyration test (i) Median test and ‘°gorov-Smirnov Test. Ifwitz test and its uses in economic data ations ? REE