Math/Stat 370: Engineering Statistics

Haijun Li
[email protected] Department of Mathematics Washington State University

Week 9

Haijun Li

Math/Stat 370: Engineering Statistics

Week 9

1 / 16

ence in the parameters of the two populations.

Comparing Means of Two Populations

Two populations respectively, unknown means 1TWO and 2 , INFERENCE FOR with, A2DIFFERENCE IN MEANS OF 2. known variances and 1 2 NORMAL DISTRIBUTIONS, VARIANCES KNOWN

Hypotheses: H0 : 1 = 2 VS H1 : 1 = 2 . More general: consider H1 : 1inferences 2 = 0 : 1 2 = 0 VS 0 . the difference In this H section we statistical on 2 Test the hypotheses with signicant level . normal distributions, where the variances 1 and 2 2 are known. Th Take a sample X1follows. 11 , X12 , . . . , as n1 from population 1. tion areX summarized Take a sample X21 , X22 , . . . , X2n2 from population 2. Two samples are independent.
Population 1
2 1

Population 2 2
2

e 10-1 Two endent populaHaijun Li

1 Sample 1: x11, x12,, x1n

2 Sample 2: x21, x22,, x2n

Math/Stat 370: Engineering Statistics

Week 9

2 / 16

Test Statistic
The sample average of the rst sample X1 =
n1 i =1 X1i

n1
n2 i =1 X2i

The sample average of the second sample X2 = E ( X 1 X 2 ) = 1 2 . V (X 1 X 2 ) =

2 1 n1

n2

2 2 n2 .

Haijun Li

Math/Stat 370: Engineering Statistics

Week 9

3 / 16

Test Statistic
The sample average of the rst sample X1 =
n1 i =1 X1i

n1
n2 i =1 X2i

The sample average of the second sample X2 = E ( X 1 X 2 ) = 1 2 . V (X 1 X 2 ) = Test statistic

2 1 n1

n2

2 2 n2 .

Z0 =

(X 1 X 2 ) 0
2 1 n1

2 2 n2

Z0 has (approximately) standard normal distribution under H0 . z is the 100 percentage point.
Haijun Li Math/Stat 370: Engineering Statistics Week 9 3 / 16

Testing Means of Two Populations, Variance Known

1

9:Desktop Folder:

Hypotheses: H0 : 1 2 = 0 VS H1 : 1 2 = 0 (two sided). Rejection region: z0 > z/2 or z0 < z/2 . Hypotheses: H0 : 1 2 = 0 VS H1 : 1 2 > 0 (upper). Rejection region: z0 > z . Hypotheses: H0 : 1 2 = 0 VS H1 : 1 2 < 0 (lower). Rejection region: z0 < z .

SES FOR A SINGLE SAMPLE

N(0,1) Critical region Acceptance region 0 ( b) z

N(0,1)

al region

Acceptance region z 0 ( c) Z0

Z0

0 is true, with critical region for (a) the two-sided alternative H1 : Figure: Percentage Points (c) the one-sided alternative H1 : 0.
Haijun Li Math/Stat 370: Engineering Statistics

0,
4 / 16

Week 9

Type II Errors
Let = 1 2 denote the true difference. Two-sided alternative, 0 0 = z/2 z/2 . 2 2 2 2 1 2 1 2 + + n1 n2 n1 n2 Upper sided alternative, 0 . = z 2 2 1 2 n1 + n2 Lower sided alternative, 0 = 1 z . 2 2 1 2 n1 + n2
Haijun Li Math/Stat 370: Engineering Statistics Week 9 5 / 16

p Values, and Sample Sizes (n1 = n2 = n)

p-value: The smallest level of signicance that leads to reject H0 . 2(1 (|z0 |)) for two-tailed test (1 (z0 )) for upper tailed test p= (z0 ) for lower tailed test Two-sided alternative, n= One sided alternative, n=
2 + 2) (z + z )2 (1 2 . ( 0 )2 2 + 2) (z/2 + z )2 (1 2 . ( 0 )2

Haijun Li

Math/Stat 370: Engineering Statistics

Week 9

6 / 16

Example
The burning rates of two different solid-fuel propellants used in aircrew escape systems are being studied. It is known that both propellants have approximately the same SD of burning rates 1 = 2 = 3 cm/s. Two samples of n1 = 20 and n2 = 20 specimens are tested, and 1 = 18.02 cm/s and x 2 = 24.37 cm/s. sample mean burning rates are x

Haijun Li

Math/Stat 370: Engineering Statistics

Week 9

7 / 16

Example
The burning rates of two different solid-fuel propellants used in aircrew escape systems are being studied. It is known that both propellants have approximately the same SD of burning rates 1 = 2 = 3 cm/s. Two samples of n1 = 20 and n2 = 20 specimens are tested, and 1 = 18.02 cm/s and x 2 = 24.37 cm/s. sample mean burning rates are x Test the hypothesis that both propellants have the same mean burning rate. Use = 0.05. Solution: Test H0 : 1 = 2 VS H1 : 1 = 2 . Since 0 = 0 and Z0 = X1 X2
2 1 n1

2 2 n2

18.02 24.37 3 1/20 + 1/20

= 6.69 < z0.025 = 1.96,

we reject H0 .

Haijun Li

Math/Stat 370: Engineering Statistics

Week 9

7 / 16

Example
The burning rates of two different solid-fuel propellants used in aircrew escape systems are being studied. It is known that both propellants have approximately the same SD of burning rates 1 = 2 = 3 cm/s. Two samples of n1 = 20 and n2 = 20 specimens are tested, and 1 = 18.02 cm/s and x 2 = 24.37 cm/s. sample mean burning rates are x Test the hypothesis that both propellants have the same mean burning rate. Use = 0.05. Solution: Test H0 : 1 = 2 VS H1 : 1 = 2 . Since 0 = 0 and Z0 = X1 X2
2 1 n1

2 2 n2

18.02 24.37 3 1/20 + 1/20

= 6.69 < z0.025 = 1.96,

we reject H0 . What is the p value of the test? Solution: Since it is a two-sided test, we have that p-value = 2(1 (| 6.69|)) = 0.
Haijun Li Math/Stat 370: Engineering Statistics Week 9 7 / 16

Example (Cont.)

Assuming equal sample sizes, what sample size should be used to ensure that = 0.01 if the true difference in means is 0.04? Solution: Since 0 = 0, = 0.04, z0.025 = 1.96 and z0.01 = 2.33, we have n=
2 + 2) (z/2 + z )2 (1 (1.96 + 2.33)2 18 2 = 208, 012. ( 0 )2 0.042

Haijun Li

Math/Stat 370: Engineering Statistics

Week 9

8 / 16

Condence Interval
1

A 100(1 )% CI for 1 2 when the variances are known: 1 x 2 z/2 x

2 1 2 1 x 2 + z/2 + 2 1 2 x n1 n2 2 1 2 + 2, n1 n2

where z/2 is the 100/2 percentage point. A 100(1 )% lower condence bound for 1 2 when the variances are known: 1 x 2 z x
2 1 2 + 2 1 2 . n1 n2

A 100(1 )% upper condence bound for 1 2 when the variances are known: 1 x 2 + z 1 2 x
2 1 2 + 2. n1 n2
Week 9 9 / 16

Haijun Li

Math/Stat 370: Engineering Statistics

Sample Size
The sample size (n1 = n2 ) required for the error 1 x 2 ) (1 2 )| E with 100(1 )% condence is given by |(x n= z/2 E
2 2 2 (1 + 2 ).

Example: Consider the propellant problem discussed above. 1 Find a 99% CI for 1 2 . Solution: Since z0.005 = 2.58, the CI is (18.02 24.37 2.58 18.02 24.37 + 2.58 or (8.8, 3.9). 33 /20 + 32 /20, 33 /20 + 32 /20),

Haijun Li

Math/Stat 370: Engineering Statistics

Week 9

10 / 16

Sample Size
The sample size (n1 = n2 ) required for the error 1 x 2 ) (1 2 )| E with 100(1 )% condence is given by |(x n= z/2 E
2 2 2 (1 + 2 ).

Example: Consider the propellant problem discussed above. 1 Find a 99% CI for 1 2 . Solution: Since z0.005 = 2.58, the CI is (18.02 24.37 2.58 18.02 24.37 + 2.58
2

33 /20 + 32 /20, 33 /20 + 32 /20),

or (8.8, 3.9). Find the sample size required for the error to be less than 1.5 cm/s with 95% condence. Solution: Since E = 1.5 and z0.025 = 1.96, we have that n = (1.96/1.5)2 18 = 30.73.
Math/Stat 370: Engineering Statistics Week 9 10 / 16

Haijun Li

Means of Two Populations, Variance Unknown

Two populations with, respectively, unknown means 1 and 2 , INFERENCE FOR A DIFFERENCE IN MEANS OF TWO 2 and 2 . unknown variances 1 2 VARIANCES KNOWN NORMAL DISTRIBUTIONS,

ence in the parameters of the two populations.

Hypotheses: H0 : 1 2 = 0 VS H1 : 1 2 = 0 . Take a sample X11 , X12 , . . . , X1n1 from population 1. In this section we consider statistical inferences on the difference Take a sample population 21 , X22 , . . . , X 2n2 from 2 2. normalX distributions, where the variances and 2 are known. Th

tion average are summarized as follows. The sample of the rst sample X 1 =

The sample average of the second sample X 2 =

Population 1 1
2

1 2 n1 i =1 X1i n1 n2 i =1 X2i n2

Population 2 2
2

e 10-1 Two endent populaHaijun Li

1 Sample 1: x11, x12,, x1n

2 Sample 2: x21, x22,, x2n

Math/Stat 370: Engineering Statistics

Week 9

11 / 16

2 2 Pooled Estimator for 1 = 2 = 2

SD:

V (X 1 X 2 ) =

2 = Sample variances: S1 n2 2 = 1 S2 i =1 (X2i X 2 n2 1

1 n2 . n1 1 i =1 (X1i n1 1 2 ) .

1 n1

X 1 )2 and

Pooled estimator (weighted average):

2 Sp = 2 + (n 1)S 2 (n1 1)S1 2 2 2 2 + (1 w )S2 . = wS1 n1 + n2 2

where w =

n1 1 n1 +n2 2

and 1 w = T0 =

n2 1 n1 +n2 2 .

Test statistic (X 1 X 2 ) 0 Sp
1 n1

1 n2

T0 has a t -distribution with n1 + n2 2 degrees of freedom under H0 .

Haijun Li Math/Stat 370: Engineering Statistics Week 9 12 / 16

hypothesis:

X 0 statistic: T2 Let t,n1 +n2 the upper 100 percentage point. 0 denote S 1n 1 Hypotheses: H0 : 1 2 = 0 VS H1 : 1 2 = 0 (two sided). ernative hypothesis Rejection criteria Rejection region: t0 > t/2,n1 +n2 2 or t0 < t/2,n1 +n2 2 . H1: Z 0 t0 t/2,n1 or t0 t/2,n1 2 Hypotheses: H : 0 1 2 = 0 VS H1 : 1 2 > 0 (upper). H1: 0 t0 t ,n1 Rejection region: ,n1 +n2 2 . H1: 0 > t,nt t0 t0 1 3 Hypotheses: H0 : 1 2 = 0 VS H1 : 1 2 < 0 (lower). Rejection region: t0 < t,n1 +n2 2 . 8 shows the location of the critical region for these situations.
tn 1 tn 1

2 2 Testing for 1 = 2 = 2

H0: 0

0 ( b)

t , n 1

t , n 1

0 ( c)

T0

n (a) =n n2 1 ution for H0: 0 with critical Figure: region for H : Z 0 , (b) H1: 0, and 1 1+
Haijun Li Math/Stat 370: Engineering Statistics Week 9 13 / 16

Example
The burning rates of two different solid-fuel propellants used in aircrew escape systems are being studied. It is known that both propellants have the same SD of burning rates 1 = 2 cm/s. Two samples of n1 = 20 and n2 = 20 specimens are tested, and sample mean burning 1 = 18.02 cm/s and x 2 = 24.37 cm/s, sample SDs are rates are x s1 = 3.11 cm/s and s2 = 2.97 cm/s. 1 Test the hypothesis that both propellants have the same mean burning rate. Use = 0.05.

Haijun Li

Math/Stat 370: Engineering Statistics

Week 9

14 / 16

Example
The burning rates of two different solid-fuel propellants used in aircrew escape systems are being studied. It is known that both propellants have the same SD of burning rates 1 = 2 cm/s. Two samples of n1 = 20 and n2 = 20 specimens are tested, and sample mean burning 1 = 18.02 cm/s and x 2 = 24.37 cm/s, sample SDs are rates are x s1 = 3.11 cm/s and s2 = 2.97 cm/s. 1 Test the hypothesis that both propellants have the same mean burning rate. Use = 0.05. Solution: Test H0 : 1 = 2 VS H1 : 1 = 2 . Since 0 = 0 and the 2 2 2 = (201)(3.11) +(201)(2.97) = 9.25, we have pooled estimator sp 20+202 T0 = X1 X2 sp
1 n1

1 n2

18.02 24.37 3.04 1/20 + 1/20

= 6.60,

which is less than t0.025,38 = 2.021. We reject H0 .

Haijun Li

Math/Stat 370: Engineering Statistics

Week 9

14 / 16

Example
The burning rates of two different solid-fuel propellants used in aircrew escape systems are being studied. It is known that both propellants have the same SD of burning rates 1 = 2 cm/s. Two samples of n1 = 20 and n2 = 20 specimens are tested, and sample mean burning 1 = 18.02 cm/s and x 2 = 24.37 cm/s, sample SDs are rates are x s1 = 3.11 cm/s and s2 = 2.97 cm/s. 1 Test the hypothesis that both propellants have the same mean burning rate. Use = 0.05. Solution: Test H0 : 1 = 2 VS H1 : 1 = 2 . Since 0 = 0 and the 2 2 2 = (201)(3.11) +(201)(2.97) = 9.25, we have pooled estimator sp 20+202 T0 = X1 X2 sp
2

1 n1

1 n2

18.02 24.37 3.04 1/20 + 1/20

= 6.60,

which is less than t0.025,38 = 2.021. We reject H0 . What is the p value of the test? Solution: The p-value 0.
Math/Stat 370: Engineering Statistics Week 9 14 / 16

Haijun Li

Condence Interval
1

A 100(1 )% CI for 1 2 when the variances are unknown: 1 x 2 t/2,n1 +n2 2 sp x 1 1 + 1 2 n1 n2 1 1 + , n1 n2

1 x 2 + t/2,n1 +n2 2 sp x

where t/2,n1 +n2 2 is the upper 100/2 percentage point.

2

A 100(1 )% lower condence bound for 1 2 when the 1 1 1 x 2 t,n1 +n2 2 sp n variances are unknown: x +n 1 2 .
1 2

A 100(1 )% upper condence bound for 1 2 when the 1 1 1 x 2 + t,n1 +n2 2 sp n variances are known: 1 2 x +n .
1 2 Math/Stat 370: Engineering Statistics Week 9 15 / 16

Haijun Li

Example

The burning rates of two different solid-fuel propellants used in aircrew escape systems are being studied. It is known that both propellants have the same SD of burning rates 1 = 2 cm/s. Two samples of n1 = 20 and n2 = 20 specimens are tested, and sample mean burning 1 = 18.02 cm/s and x 2 = 24.37 cm/s, sample SDs are rates are x s1 = 3.11 cm/s and s2 = 2.97 cm/s. Find a 95% CI for 1 2 .

Haijun Li

Math/Stat 370: Engineering Statistics

Week 9

16 / 16

Example

The burning rates of two different solid-fuel propellants used in aircrew escape systems are being studied. It is known that both propellants have the same SD of burning rates 1 = 2 cm/s. Two samples of n1 = 20 and n2 = 20 specimens are tested, and sample mean burning 1 = 18.02 cm/s and x 2 = 24.37 cm/s, sample SDs are rates are x s1 = 3.11 cm/s and s2 = 2.97 cm/s. Find a 95% CI for 1 2 . Solution: Since t0.025,38 = 2.021, the CI is (18.02 24.37 2.021 3.04 1/20 + 1/20, 18.02 24.37 + 2.021 3.04 1/20 + 1/20), or (8.29, 4.41).

Haijun Li

Math/Stat 370: Engineering Statistics

Week 9

16 / 16