Figure 1: Dierent Statistical Tests for dierent scenarios This is a quick compilation of relevant formulae for the end

trimester syllabus. For each section, the shaded items, e.g. P. 406 Ex. 8.2 , refer to exercises from Statistics for Management by Levin and Rubin, that you should be familiar with.

Mean, Standard Deviation

Combined Arithmetic Mean of k sets of data, with the ith set having ni elements
k

ni x i x =
i=1 k

ni
i=1

For data x1 , x2 , . . . , xn , the variance is dened as 2 = It is also written as 2 = 1 n 1 n

n

(xi x )2
i=1 n

xi 2 x 2
i=1

In real life situations, whenever data is obtained as a sample, the variance of the sample is
n n

(xi x 2 ) n1 Population Standard Deviation = 2 Sample Standard Deviation = s2 = s s2 =

i=1

x2 2 i nx =
i=1

n1

Hypothesis Testing - One Sample Tests

x = n x z= x

P. 406 Ex. 8.2 .

Q 8.2, 8-12 A grocery store has specially packaged oranges and has claimed a bag of oranges will yield 2.5 quarts of juice. After randomly selecting 42 bags, a stocker found the average juice production per bag to be 2.2 2

quarts. Historically, we know the population standard deviation is 0.2 quart. Using this sample and a decision criterion of 2.5 standard errors, could we conclude the stores claims are correct? A 8.2, 8-12 H0 : = 2.5 H1 : < 2.5 n = 42, x = 2.2, = 0.2 x 2 . 2 2 .5 = = -9.72 0.2
n 42

zobs < -2.5 (2.5 standard errors left of H0 ) So reject H0 . Stores claims are not correct. . Condence Interval: H0 zcrit x One tailed test of means P. 422 Ex. 8.4 .

Q 8.4, 8-28 Generally Electric has developed a new bulb whose design specications call for a light output of 960 lumens compared to an earlier model that produced only 750 lumens. The companys data indicate that the standard deviation of light output for this type of bulb is 18.4 lumens. From a sample of 20 new bulbs, the testing committee found an average light output of 954 lumens per bulb. At a 0.05 signicance level, can Generally Electric conclude that its new bulb is producing the specied 960 lumen output? A 8.4, 8-28 H0 : = 960 H1 : = 960 n = 20, x = 954, = 18.4, = 0.05 x 954960 = = -1.4583 18.4
n

zcrit = 1.96 Since -1.96 < zobs = 1.4583 < +1.96, we do not reject H0 Bulb is producing the required 960 lumen. .

20

Power P. 426 Ex. 8.5

2.1

Hypothesis Test for Proportion

p = pH0 qH0 n p pH0 z= p

P. 431 Ex. 8.6 .

Q 8.6, 8-40 Feronetics specializes in the use of gene-splicing techniques to produce new pharmaceutical compounds. It has recently developed a nasal spray containing interferon, which it believes will limit the transmission of the common cold within families. In the general population, 15.1 percent of all individuals will catch a rhinovirus-caused cold once another family member contracts such a cold. The interferon spray was tested on 180 people, one of whose family members subsequently contracted a rhinovirus-caused cold. Only 17 of the test subjects developed similar colds. (a) At a signicance level of 0.05, should Feronetics conclude that the new spray eectively reduces transmission of colds? (b) What should it consider at = 0.02? A 8.6, 8-40 H0 : pH0 = 0.151 H1 : pH0 < 0.151 17 = 0.0944, = 0.05 n = 180, p = 180 p =
p pH 0 p pH0 qH0 0.849 = 0.151 = n 180 0.151 = 0.0944 = -2.1209 0.0266

0.0266

(a) zcrit at 0.05 = 1.645 Since zobs = 2.1209 < -1.645, we reject H0 . The new spray eectively reduces transmission of cold. (b) zcrit at 0.02 = 2.055 Since zobs = 2.1209 < -2.055, we still reject H0 . The new spray eectively reduces transmission of cold. .

2.2

Hypothesis Test for Means when is not known

x = n =s x H 0 t= x P. 436 Ex. 8.7 .

Q 8.7, 8-48 The data-processing department at a large life insurance company has installed new color video display terminals to replace the monochrome units it previously used. The 95 operators trained to use the new machines averaged 7.2 hours before achieving a satifactory level of performance. Their sample variance was 16.2 squared hours. Long experience with operators on the old monochrome terminals showed that they averaged 8.1 hours on the machines before their performances were satisfactory. At the 0.01 signicance level, should the supervisor of the department conclude that the new terminals are easier to learn to operate? A 8.7, 8-48 H0 : = 8.1 H1 : < 8.1 n = 95, x = 7.2, = 16.2, = 0.01 8 .1 zobs = 7.2 = -2.179 16.2
95

zcrit at 0.01 = 2.33 Since -2.33 < zobs = 2.179 < +2.055, we do not reject H0 . The new terminals are not signicantly easier to operate. .

3
3.1

Hypothesis Testing - Two Sample Tests

Tests for dierence between Means
x 1 x 2 =
2 1 2 + 2 n1 n2

=s

 x 1 x 2 = z=

2 1 2 + 2 n1 n2

( x1 x 2 ) (1 2 )H0 x 1 x 2

Large Sample Sizes P. 459 Ex. 9.2 .

Q 9.2, 9-4 A sample of 32 money-market mutual funds was chosen on January 1, 1996, and the average annual rate of return over the past 30 days was found to be 3.23 percent, and the sample standard deviation was 0.51 percent. A year earlier, a sample of 38 money-market funds showed an average rate of return of 4.36 percent. Is it reasonable to conclude (at = 0.05) that the money-market interest rates declined during 1995? A 9.2, 9-4 Jan 1, 1996 Jan 1, 1995 = 0.05, n2 = 32, x 2 = 3.23%, s2 = 0.51 n1 = 38, x 1 = 4.36%, s1 = 0.84 H0 : 2 1 = 0, H1 : 2 1 < 0 zobs = x 2 x 1
2 1 n1 2 2 n2

3.23 4.36
0.842 38

0.512 32

= 6.9155

zcrit (at = 0.05) is 1.645 Since zobs = -6.9155 < zobs = -1.645, we reject H0 Conclusion: Rates have declined. .

Small Sample Sizes Pooled estimate of 2 s2 p =

2 (n1 1)s2 1 + (n2 1)s2 n1 + n2 2

x 1 x 2 = sp t=

1 1 + n1 n2

( x1 x 2 ) (1 2 )H0 x 1 x 2

degrees of freedom for t distribution = n1 + n2 2 P. 465 Ex. 9.3 .

Q 9.3, 9-8 A credit-insurance organization has developed a new high-tech method of training new sales personnel. The company sampled 16 employees who were trained the original way and found average daily sales to be $688 and the sample standard deviation was $32.63. They also sampled 11 employees who were trained using the new method and found average daily sales to be $706 and the sample standard deviation was $24.84. At = 0.05, can the company conclude that average daily sales have increased under the new plan? A 9.3, 9-8 n1 = 16, x 1 = 688, s1 = 32.63 n2 = 11, x 2 = 706, s2 = 24.84 = 0.05, H0 : 2 1 = 0, H1 : 2 1 > 0 2 2 Pooled estimate of , sp
2 (n1 1)s2 15(32.63)2 + 10(24.84)2 1 + (n2 1)s2 = = 885.64 n1 + n2 2 16 + 11 2 sp = 885.64 = 29.76

x 2 x 1 = 29.76

1 1 + = 11.656 16 11

zobs =

( x2 x 1 ) (2 1 ) 706 688 = 1.544 = x 11.656 2 x 1

tcrit (25 df at = 0.1) is 1.708 Since tobs = 1.544 < tcrit = 1.708, we do not reject H0 Conclusion: Average sales has not increased signicantly under the new plan. .

3.2

Testing dierence between Means (dependent samples)

s x = = n n 7

t= P. 473 Ex. 9.4 .

x H0 x

Q 9.4, 9-14 The data below are a random sample of 9 rms chosen from the Digest of Earnings Reports in The Wall Street Journal on February 6, 1992: (a) Find the mean change in earnings per share between 1991 and 1992 (b) Find the standard deviation of the change and the standard error of the mean. (c) Were the average earnings per share dierent in 1991 and 1992? Test at = 0.02. Firm 1991 earnings 1992 earnings 1 1.38 2.48 2 1.26 1.50 3 3.64 4.59 4 3.50 3.06 5 2.47 2.11 6 3.21 2.80 7 1.05 1.59 8 1.98 0.92 9 2.72 0.47

A 9.4, 9-14 = 0.02, H0 : dif f = 0, H1 : dif f = 0 The dierences in earnings are as follows: Firm 1 2 3 4 5 6 Dierence in earnings 1.10 0.24 0.95 -0.44 -0.36 -0.41 x = 1.69, x2 = 9.1391, x = 0.18777, n=9 s= x2 9 x2 = 1.1027 = 1.05 91
1 .05 9

7 0.54

8 -1.06

9 -2.25

(a) Mean change, x = -0.1878 (b) s = 1.05, (c) tobs =

x H 0
dif f n

sx = =

= 0.35 = 0.536

0.18777 0.35

tcrit (8df, = 0.02) = 2.896 Since tobs = 0.536 < tcrit = 2.896, we do not reject H0 Conclusion: the average earnings did not change signicantly. .

3.3

Tests for dierence between proportions - Large Samples

p = p 1 p 2 = p 1 p 2 = pq n p 1 q 1 p2 q 2 + n1 n2 2 q 2 p 1 q 1 p + n1 n2

p 1 , q 1 , p 2 and q 2 are the sample proportions p = n1 p 1 + n2 p 2 n1 + n2 p q 1 1 + n1 n2

p 1 p 2 = z= P. 482 Ex. 9.5 .

( p1 p 2 ) (p1 p2 )H0 p 1 p 2

Q 9.5, 9-22 A coal-red power plant is considering two dierent systems for pollution abatement. The rst system has reduced the emissionof pollutants to acceptable levels 68 percent of the time, as determined from 200 air samples. The second, more expensive system has reduced the emission of pollutants to acceptable levels 76 percent of the time, as determined from 250 air samples. If the expensive system is signicantly more eective than the inexpensive system in reducing pollutants to acceptable levels, then the management of the power plant will install the expensive system. Which system will be installed if management uses a signicance level of 0.02 in making its decision? A 9.5, 9-22 p 1 = 0.68, p 2 = 0.76, = 0.02, n1 = 200 n2 = 250 H0 : 2 1 = 0, p 2 p 1 = zobs =

H1 : 2 1 > 0

p 1 q 1 p2 q 2 + = 0.0426 n1 n2

( p2 p 1 ) (2 1 )H0 = 1.8765 p 2 p 1 9

zcrit (at = 0.02) = 2.055 Since zobs = 1.8765 < zcrit = 2.055, we do not reject H0 Conclusion: the 2nd system is not signicantly more eective . P. 488 Ex. 9.6

2 Tests

Chi-Square Tests (fobs fexp )2 2 = fexp where fobs = observed cell frequency, fexp = expected cell frequency 2 degrees of freedom = (rows - 1)(columns - 1) ColumnT otal Expected frequency of a cell, fexp = RowT otal GrandT otal 2 Test of Independence : P. 581 Ex. 11.2 .

Q 11.2, 11-7, 11-8 An advertising rm is trying to determine the demographics for a new product. They have randomly selected 75 people in each of 5 dierent age groups nand introduced the product to them. The results of the survey are given below: Age Group Future Activity 18-29 30-39 40-49 50-59 60-69 Purchase Frequently 12 18 17 22 32 Seldom Purchase 18 25 29 24 30 Never Purchase 45 32 29 29 13 Develop a table of observed and expected frequencies for this problem. 11-8 (a) Calculate the sample 2 value 11-8 (b) State the null and alternate hypotheses 11-8 (c) If the level of signicance is 0.01, should the null hypothesis be rejected? A 11.2, 11-7, 11-8 The table of observed and expected frequencies is

10

Future Activity Purchase Frequently

O
CT E ( RT T ot. ) (O E )2 (O E )2 E

18-29 12 20.2 67.24 3.33 18 25.2 51.84 2.06 45 29.6 237.16 8.01 75

30-39 18 20.2 4.84 0.24 25 25.2 0.04 0.00 32 29.6 5.76 0.19 75
2

Age Group 40-49 17 20.2 10.24 0.51 29 25.2 14.44 0.57 29 29.6 0.36 0.01 75

50-59 22 20.2 3.24 0.16 24 25.2 1.44 0.06 29 29.6 0.36 0.01 75

60-69 32 20.2 139.24 6.89 30 25.2 23.04 0.91 13 29.6 275.56 9.31 75

Row Marginal 101

Seldom Purchase

O E (O E )
(O E )2 E CT ( RT T ot. ) 2

126

Never Purchase

O CT E ( RT T ot. ) (O E )2
(O E )2 E

148

Column Marginals

375

(O E ) 2 2 = 32.36 crit,0.01,42=8df = 20.09; obs = E null hypothesis is rejected; the purchasing habits is dependent on age group

2 Test of Goodness of t: P. 587 Ex. 11.3 .

Q 11.3, 11-17 The computer coordinator of a business school believes the amount of time a graduate student spends reading and writing e-mail each weekday is normally distributed with mean = 14 and standard deviation = 5. In order to examine this belief, the coordinatorcollected data one Wednesday, recording the amount of time in minutes each graduate student spent checking e-mail. Using a chi-square goodness-of-t test on these data, what would you conclude about the distribution of e-mail times? (Use a 0.05 signicance level and clearly state your hypothesis.)(Hint: Use ve equally probable intervals.) 8.2 7.4 9.6 12.8 22.4 6.2 8.7 9.7 12.4 10.6 1.2 18.6 3.3 15.7 18.4 12.4 15.9 19.4 12.8 20.4 12.3 11.3 10.9 18.4 14.3 16.2 6.7 13.9 18.3 19.2 14.3 14.9 16.7 11.3 18.4 18.8 20.4 12.4 18.1 20.1 11

A 11.3, 11-17 Following is the table of observed and expected frequencies Observed z (upper X N(14,5) Class 0 - <5 5 - <10 10 - <15 15 - <20 20 - <25 Frequency 2 7 14 13 4 40 class lim.) 514 =-1.8 5 1014 =-0.8 5 1514 =0.2 5 2014 5 =1.2 2514 5 =2.2 P(X<z) 0.0359 0.2119 0.5793 0.8849 0.9861
2

Expected Frequency (E) 40*.0359=1.44 40*.1760=7.04 40*.3674=14.70 40*.3056=12.22 40*.1012=4.05

(O E ) 2 E

P(X class) 0.0359 0.1760 0.3674 0.3056 0.1012

0.2178 0.0002 0.0333 0.0498 0.0006

(O E ) 2 2 = 0.3017 crit,0.05,4df = 9.488; obs = E null hypothesis is accepted; the e-mail times N(14,5)

Analysis of Variance (ANOVA)

)2 nj ( xj x k1

Between Columns (Groups) Variance

2 = b

Within Columns (Groups) Variance

2 w =

nj 1 2 s nT k j

where k = number of groups, nT is the Grand Total, nj = number of items in the j th group
th group F = and s2 j = sample variance within the j Degrees of Freedom of F = (k 1, nT k ) P. 603 Ex. 11.4 .
2 b 2 w

Q 11.4, 11-26 A study compared the number of hours of relief provided by ve different brands of antacid administered to 25 dierent people, each with stomach acid considered strong. The results are given below: 12

Brand

A 4.4 4.6 4.5 4.1 3.8

B 5.8 5.2 4.9 4.7 4.6

C 4.8 5.9 4.9 4.6 4.3

D 2.9 2.7 2.9 3.9 4.3

E 4.6 4.3 3.8 5.2 4.4

(a) Compute the mean number of hours of relief for each brand and determine the grand mean. (b) Estimate the population variance using the between-column variance. (c) Estimate the population variance using the within-column variance computed from the variance within the samples. (d) Calculate the F ratio. At the 0.05 level of signicance, do the brands produce signicantly dierent amounts of relief to people with strong stomach acid? A 11.4, 11-26 Computation table: Brand A 4.4 4.6 4.5 4.1 3.8 21.4 4.28 B 5.8 5.2 4.9 4.7 4.6 25.2 5.04 0.932 0.404496 2.02248 C 4.8 5.9 4.9 4.6 4.3 24.5 4.90 1.460 0.246016 1.23008 D 2.9 2.7 2.9 3.9 4.3 16.7 3.34 2.032 1.132096 5.66048 E 4.6 4.3 3.8 5.2 4.4 22.3 4.46 1.032 0.003136 0.01568

X X = 4.404 Grand Mean X )2 (A) (Xi X 0.428 2 X SSB X 0.015376 (B) = n SSB = 5.SSB 0.07688 2 = b (B ) / 4 = 2.2514 2 = (A) / (25 - 5) = 0.2942
w

.2514 Fobs (df 4, 20) = 2 0.2942 = 7.6526 Fcrit (df 4, 20; 0.05 signicance) = 2.8661; Since Fobs > Fcrit reject the null hypothesis of equality of means Conclusion: the brands produce signicantly dierent amounts of relief to people with strong stomach acid.

13

6
6.1

Correlation and Regression

Estimation using Regression line
b= Y XY nX 2 X nX bX a=Y

Standard Error of the Regression Estimate se = P. 671 Ex. 12.2 . Y2a Y b n2 XY

Q 12.2, 12-18 A study by the Atlanta, Georgia, Department of Transportation on the eect of bus-ticket prices on the number of passengers produced the following results: Ticket price (cents) 25 30 35 40 45 50 55 60 Passengers per 100 miles 800 780 780 660 640 600 620 620 (a) Plot these data. (b) Develop the estimating equation that best describes these data. (c) Predict the number of passengers per 100 miles if the ticket prices were 50 cent. Use a 95 percent approximate prediction interval. A 12.2, 12-18 From the data provided, we get X = 340; Y = 5500 X 2 = 15500; Y 2 = 3830800 = 42.5; = 687.5; X Y XY = 227200 b= Y XY nX 6550 = = 6.238 2 2 1050 X nX

bX = 687.5 + 6.238 42.5 = 952.615 a=Y Y = 952.615 - 6.238 X P ASSEN GERS = 952.615 6.238.T ICKET P RICE

14

P ASSEN GERS T icketP rice=50 = 640.715 Y b XY 8691.1 = sqrt = 38.06 n2 6 95% prediction interval for number of passengers at a ticket price of 50 cents is P ASSEN GERS T icketP rice=50 se tcrit,0.05 = 640.715 38.06 2.447 = 640.715 93.132 se = . Y2a

6.2

Correlation

Coecient of Determination = r2 r2 = Y XY nX 2 )( Y 2 nY 2) X 2 nX r2

Correlation Coecient r = P. 685 Ex. 12.3 .

Q 12.3, 12-32 Zippy Cola is studying the eect of its latest advertising campaign. People chosen at random were called and asked how many cans of Zippy Cola they had bought in the past week and how many Zippy Cola advertisements they had either read or seen in the past week. X (number of ads) 3 7 4 2 0 4 1 2 Y (cans purchased) 11 18 9 4 7 6 3 8 (a) Develop the estimating equation that best ts the data. (b) Calculate the sample coecient of determination and the sample coecient of correlation. A 12.3, 12-32 From the data provided, the following summary statistics emerge: = 2.875 X = 23; X 2 = 99; X 2 = 8.25 Y = 66; Y = 700; Y XY = 246 b= Y XY nX 56.25 2 = 32.875 = 1.711 X 2 nX 15

 = bX = 3.33 a=Y P U RCHASES = 3.33 + 1.71 ADS r2 = Y XY nX 96.186 = 2 2 2 2 155.5 X nX )( Y nY )

2 = 0.6186 Coecient of Determination, r Coecient of Correlation, r = 0.6186 = 0.7865 r has a positive sign because b, the slope of the regression equation, is +ve

6.3

Inferences about the slope of the regression line

se 2 X 2 nX b BH0 sb

Standard Error of the Regression Slope, b sb =

t=

Condence Interval of the regression coecient = b tsb P. 691 Ex. 12.4 .

Q 12.4, 12-38 In 1969, a government health agency found that in a number of counties, the relationship between smokers and heart-disease fatalities per 100,000 population had a slope of 0.08. A recent study of 18 counties produced a slope of 0.147 and a standard error of the regression slope coecient of 0.032. (a) Construct a 90 percent condence interval estimate of the slope of the true regression line. Does the result from this study indicate that the true slope has changed? (b) Construct a 99 percent condence interval estimate of the slope of the true regression line. Does the result from this study indicate that the true slope has changed?

16

 A 12.4, 12-38 b = 0.08; b = 0.147; sb = 0.032; t1 6df, 0.1 = 1.746; t1 6df, 0.01 = 2.921

n = 18

(a) 90% condence interval for b is b sb tcrit = 0.147 0.032 1.746 = 0.147 0.05587 = (0.091, 0.203) Since, 0.08, the earlier population slope, lies outside this range, we conclude that the slope has changed at a 90% level of signicance (b) 99% condence interval for b is b sb tcrit = 0.147 0.032 2.921 = 0.147 0.09347 = (0.054, 0.240) Since, 0.08, the earlier population slope, lies inside this range, we conclude that the slope has not changed at a 99% level of signicance .

7
7.1

Non Parametric Tests

Sign Test (Paired Data)
p = pH0 qH0 n p pH0 z= p

7.2

Mann-Whitney U Test
U = n1 n2 + n1 (n1 + 1) R1 2 n1 n2 U = 2

U =

n1 n2 (n1 + n2 + 1) 12 U U z= U

P. 809 Ex. 14.3, Problems 14-14, 14-17, 14-18, 14-20 .

17

Q 14.3, 14-14 Test the hypothesis of no dierence between the ages of male and female employees of a certain company using the Mann-Whitney U test for the sample data. Use the 0.10 level of signicance. Males Females 31 44 25 30 38 34 33 47 42 35 40 32 44 35 26 47 43 48 35 34

A 14.3, 14-14 H0 : m = f ; H1 : m = f ; = 0.1 The following rankings emerge from the given data Age Gender Rank 25 M 1 26 M 2 30 F 3 31 M 4 32 F 5 33 M 6 34 F 7.5 34 F 7.5 35 M 10 35 F 10 35 F 10 38 M 12 40 M 13 42 M 14 43 M 15 44 M 16.5 44 F 16.5 47 F 18.5 47 F 18.5 48 F 20 nm = 10; nf = 10; Rm = 93.5 Rf = 116.5

18

U = nm nf +

nm (nm + 1) Rm 2 (10)(11) = (10)(10) + 93.5 2 = 100 + 55 93.5

= 61.5
nm nf = (10)(10) 2 2 nm nf (nm +nf +1) 12

Mean of the sampling distribution of U , U = Standard error of the U statistic, U = U =

= 50

(10)(10)(10 + 10 + 1) 12 2100 = 12 = 175 = 13.23

U .550 zobs = U = 61 U 13.23 = 0.8692 Since zL = 1.645 < zobs < zU , we conclude that there is no dierence between the ages. .

7.3

Kruskal-Wallis Test
K= 12 n(n + 1)
k j =1 2 Rj 3(n + 1) nj

where k = number of groups K 2 df (n 1) P. 809 Ex. 14.3, Problems 14-15, 14-16, 14-19 .

Q 14.3, 14-16 A mail-order gift company has the following sample data on dollar sales, separated according to how the order was paid. Test the hypothesis that there is no dierence in the dollar amount of orders paid 19

for by cash, by check, or by credit card. with a 0.05 level of signicance. Credit-card orders 78 64 75 45 Check orders 110 70 53 51 Cash orders 90 68 70 54

Use the Kruskal-Wallis test 82 61 74 69 68 65 60 59 =

A 14.3, 14-16 H0 : cc = ck = cs ; H1 : the means are not all equal; 0.05 The following rankings emerge from the given data Dollar amount Mode of of Orders paid Payment Rank 45 CC 1 51 CK 2 53 CK 3 54 CS 4 59 CS 5 60 CC 6 61 CK 7 64 CC 8 65 CS 9 68 CK 10.5 68 CS 10.5 69 CC 12 70 CS 13.5 70 CK 13.5 74 CS 15 75 CC 16 78 CC 17 82 CC 18 90 CS 19 110 CK 20 k=3 nCC = 7; nCK = 6; nCS = 7; RCC = 78 RCK = 56 RCS = 76

20

12 K= n(n + 1)

k j =1

2 Rj 3(n + 1) nj

12 6084 3136 5776 = ( + + ) 3(20 + 1) (20)(21) 7 6 7 = 0.02857(869.14 + 522.67 + 825.14) 63 = 0.02857 2216.95 63 = 63.34 63 = 0.34 2 (2 df, signicance 0.05) = 5.991 Since K = 0.34 < 2 crit = 5.991, we do not reject H0 Conclusion: The average amounts paid by the three methods are not significantly dierent. .

7.4

One Sample Run Test

r = 2n1 n2 +1 n1 + n2

where n1 = sample size of one category n2 = sample size of the other category r = number of runs r = 2n1 n2 (2n1 n2 n1 n2 ) (n1 + n2 )2 (n1 + n2 1) z= P. 817 Ex. 14.4 . r r r

Q 14.4, 14-26 The N ewsandClarion kept a record of the gender of people who called the circulation oce to complain about delivery problems with the Sunday paper. For a recent Sunday, these data were as follows: 21

M,F,F,F,M,M,F,M,F,F,F,F,M,M,M,F,M,F,M,F,F,F,F,M,M,M,M,M Using the 0.05 level of signicance, test this sequence for randomness. Is there anything about the nature of this problem that would cause you to believe that such a sequence would not be random? A 14.4, 14-26 From the run sequence of geneders provided in the question: n1 = 14, n2 = 14, r = 13, = 0.05 2n 1 n 2 +1 n1 + n2 2 14 14 +1 = 14 + 14 = 15

r =

r = =

2n1 n2 (2n1 n2 n1 n2 ) (n1 + n2 )2 (n1 + n2 1)

2.14.14.(2.14.14 28) 282 .27 14.13 = sqrt 27 = sqrt6.74 = 2.6 r r r 13 15 = 2.6 = 0.77

zobs =

zcrit0.05 = 1.96 Since zobs < |zcrit0.05 | = 1.96, we do not reject H0 Conclusion: the sequence appears to be random. .

22