![Page 1: Math 362: Mathmatical Statistics IImath.emory.edu/~lchen41/teaching/2020_Spring/Chapter-7_full.pdf · Gosset worked at the Guinness Brewery in Dublin, Ireland, and was interested](https://reader036.vdocuments.site/reader036/viewer/2022071005/5fc1e109d75c9e182e3dd73a/html5/thumbnails/1.jpg)
Math 362: Mathmatical Statistics II
Emory UniversityAtlanta, GA
Last updated on February 27, 2020
2020 Spring
0
![Page 2: Math 362: Mathmatical Statistics IImath.emory.edu/~lchen41/teaching/2020_Spring/Chapter-7_full.pdf · Gosset worked at the Guinness Brewery in Dublin, Ireland, and was interested](https://reader036.vdocuments.site/reader036/viewer/2022071005/5fc1e109d75c9e182e3dd73a/html5/thumbnails/2.jpg)
Chapter 7. Inference Based on The NormalDistribution
1
![Page 3: Math 362: Mathmatical Statistics IImath.emory.edu/~lchen41/teaching/2020_Spring/Chapter-7_full.pdf · Gosset worked at the Guinness Brewery in Dublin, Ireland, and was interested](https://reader036.vdocuments.site/reader036/viewer/2022071005/5fc1e109d75c9e182e3dd73a/html5/thumbnails/3.jpg)
Plan
§ 7.1 Introduction
§ 7.2 Comparing Y−µσ/√
n and Y−µS/√
n
§ 7.3 Deriving the Distribution of Y−µS/√
n
§ 7.4 Drawing Inferences About µ
§ 7.5 Drawing Inferences About σ2
2
![Page 4: Math 362: Mathmatical Statistics IImath.emory.edu/~lchen41/teaching/2020_Spring/Chapter-7_full.pdf · Gosset worked at the Guinness Brewery in Dublin, Ireland, and was interested](https://reader036.vdocuments.site/reader036/viewer/2022071005/5fc1e109d75c9e182e3dd73a/html5/thumbnails/4.jpg)
Chapter 7. Inference Based on The Normal Distribution
§ 7.1 Introduction
§ 7.2 Comparing Y−µσ/√
n and Y−µS/√
n
§ 7.3 Deriving the Distribution of Y−µS/√
n
§ 7.4 Drawing Inferences About µ
§ 7.5 Drawing Inferences About σ2
3
![Page 5: Math 362: Mathmatical Statistics IImath.emory.edu/~lchen41/teaching/2020_Spring/Chapter-7_full.pdf · Gosset worked at the Guinness Brewery in Dublin, Ireland, and was interested](https://reader036.vdocuments.site/reader036/viewer/2022071005/5fc1e109d75c9e182e3dd73a/html5/thumbnails/5.jpg)
§ 7.1 Introduction
(1777-1855) (1749-1827)
4
![Page 6: Math 362: Mathmatical Statistics IImath.emory.edu/~lchen41/teaching/2020_Spring/Chapter-7_full.pdf · Gosset worked at the Guinness Brewery in Dublin, Ireland, and was interested](https://reader036.vdocuments.site/reader036/viewer/2022071005/5fc1e109d75c9e182e3dd73a/html5/thumbnails/6.jpg)
https://en.wikipedia.org/wiki/Normal_distribution
5
![Page 7: Math 362: Mathmatical Statistics IImath.emory.edu/~lchen41/teaching/2020_Spring/Chapter-7_full.pdf · Gosset worked at the Guinness Brewery in Dublin, Ireland, and was interested](https://reader036.vdocuments.site/reader036/viewer/2022071005/5fc1e109d75c9e182e3dd73a/html5/thumbnails/7.jpg)
Test for normal parameters (one sample test)
Let Y1, · · · ,Yn be a random sample from N(µ, σ2).
Prob. 1 Find a test statistic Λ in order to test H0 : µ = µ0 v.s. H1 : µ 6= µ0.
When σ2 is known: Λ =Y − µ0
σ/√
n∼ N(0, 1)
When σ2 is unknown: Λ =? Λ?=
Y − µ0
s/√
n∼ ?
Prob. 2 Find a test statistic Λ in order to test H0 : σ2 = σ20 v.s. H1 : σ2 6= σ2
0 .
6
![Page 8: Math 362: Mathmatical Statistics IImath.emory.edu/~lchen41/teaching/2020_Spring/Chapter-7_full.pdf · Gosset worked at the Guinness Brewery in Dublin, Ireland, and was interested](https://reader036.vdocuments.site/reader036/viewer/2022071005/5fc1e109d75c9e182e3dd73a/html5/thumbnails/8.jpg)
Test for normal parameters (one sample test)
Let Y1, · · · ,Yn be a random sample from N(µ, σ2).
Prob. 1 Find a test statistic Λ in order to test H0 : µ = µ0 v.s. H1 : µ 6= µ0.
When σ2 is known: Λ =Y − µ0
σ/√
n∼ N(0, 1)
When σ2 is unknown: Λ =? Λ?=
Y − µ0
s/√
n∼ ?
Prob. 2 Find a test statistic Λ in order to test H0 : σ2 = σ20 v.s. H1 : σ2 6= σ2
0 .
6
![Page 9: Math 362: Mathmatical Statistics IImath.emory.edu/~lchen41/teaching/2020_Spring/Chapter-7_full.pdf · Gosset worked at the Guinness Brewery in Dublin, Ireland, and was interested](https://reader036.vdocuments.site/reader036/viewer/2022071005/5fc1e109d75c9e182e3dd73a/html5/thumbnails/9.jpg)
Test for normal parameters (one sample test)
Let Y1, · · · ,Yn be a random sample from N(µ, σ2).
Prob. 1 Find a test statistic Λ in order to test H0 : µ = µ0 v.s. H1 : µ 6= µ0.
When σ2 is known: Λ =Y − µ0
σ/√
n∼ N(0, 1)
When σ2 is unknown: Λ =? Λ?=
Y − µ0
s/√
n∼ ?
Prob. 2 Find a test statistic Λ in order to test H0 : σ2 = σ20 v.s. H1 : σ2 6= σ2
0 .
6
![Page 10: Math 362: Mathmatical Statistics IImath.emory.edu/~lchen41/teaching/2020_Spring/Chapter-7_full.pdf · Gosset worked at the Guinness Brewery in Dublin, Ireland, and was interested](https://reader036.vdocuments.site/reader036/viewer/2022071005/5fc1e109d75c9e182e3dd73a/html5/thumbnails/10.jpg)
Test for normal parameters (one sample test)
Let Y1, · · · ,Yn be a random sample from N(µ, σ2).
Prob. 1 Find a test statistic Λ in order to test H0 : µ = µ0 v.s. H1 : µ 6= µ0.
When σ2 is known: Λ =Y − µ0
σ/√
n∼ N(0, 1)
When σ2 is unknown: Λ =? Λ?=
Y − µ0
s/√
n∼ ?
Prob. 2 Find a test statistic Λ in order to test H0 : σ2 = σ20 v.s. H1 : σ2 6= σ2
0 .
6
![Page 11: Math 362: Mathmatical Statistics IImath.emory.edu/~lchen41/teaching/2020_Spring/Chapter-7_full.pdf · Gosset worked at the Guinness Brewery in Dublin, Ireland, and was interested](https://reader036.vdocuments.site/reader036/viewer/2022071005/5fc1e109d75c9e182e3dd73a/html5/thumbnails/11.jpg)
Test for normal parameters (one sample test)
Let Y1, · · · ,Yn be a random sample from N(µ, σ2).
Prob. 1 Find a test statistic Λ in order to test H0 : µ = µ0 v.s. H1 : µ 6= µ0.
When σ2 is known: Λ =Y − µ0
σ/√
n∼ N(0, 1)
When σ2 is unknown: Λ =? Λ?=
Y − µ0
s/√
n∼ ?
Prob. 2 Find a test statistic Λ in order to test H0 : σ2 = σ20 v.s. H1 : σ2 6= σ2
0 .
6
![Page 12: Math 362: Mathmatical Statistics IImath.emory.edu/~lchen41/teaching/2020_Spring/Chapter-7_full.pdf · Gosset worked at the Guinness Brewery in Dublin, Ireland, and was interested](https://reader036.vdocuments.site/reader036/viewer/2022071005/5fc1e109d75c9e182e3dd73a/html5/thumbnails/12.jpg)
Test for normal parameters (one sample test)
Let Y1, · · · ,Yn be a random sample from N(µ, σ2).
Prob. 1 Find a test statistic Λ in order to test H0 : µ = µ0 v.s. H1 : µ 6= µ0.
When σ2 is known: Λ =Y − µ0
σ/√
n∼ N(0, 1)
When σ2 is unknown: Λ =? Λ?=
Y − µ0
s/√
n∼ ?
Prob. 2 Find a test statistic Λ in order to test H0 : σ2 = σ20 v.s. H1 : σ2 6= σ2
0 .
6
![Page 13: Math 362: Mathmatical Statistics IImath.emory.edu/~lchen41/teaching/2020_Spring/Chapter-7_full.pdf · Gosset worked at the Guinness Brewery in Dublin, Ireland, and was interested](https://reader036.vdocuments.site/reader036/viewer/2022071005/5fc1e109d75c9e182e3dd73a/html5/thumbnails/13.jpg)
Prob. 1 Find a test statistic for H0 : µ = µ0 v.s. H1 : µ 6= µ0, with σ2 unknown
Sol. Composite-vs-composite test with:
ω =
(µ, σ2) : µ = µ0, σ2 > 0
Ω =
(µ, σ2) : µ ∈ R, σ2 > 0
The MLE under the two spaces are:
ωe = (µe, σ2e) : µe = µ0 and σ2
e =1n
n∑i=1
(yi − µ0)2 (Under ω)
Ωe = (µe, σ2e) : µe = y and σ2
e =1n
n∑i=1
(yi − y)2 (Under Ω)
7
![Page 14: Math 362: Mathmatical Statistics IImath.emory.edu/~lchen41/teaching/2020_Spring/Chapter-7_full.pdf · Gosset worked at the Guinness Brewery in Dublin, Ireland, and was interested](https://reader036.vdocuments.site/reader036/viewer/2022071005/5fc1e109d75c9e182e3dd73a/html5/thumbnails/14.jpg)
Prob. 1 Find a test statistic for H0 : µ = µ0 v.s. H1 : µ 6= µ0, with σ2 unknown
Sol. Composite-vs-composite test with:
ω =
(µ, σ2) : µ = µ0, σ2 > 0
Ω =
(µ, σ2) : µ ∈ R, σ2 > 0
The MLE under the two spaces are:
ωe = (µe, σ2e) : µe = µ0 and σ2
e =1n
n∑i=1
(yi − µ0)2 (Under ω)
Ωe = (µe, σ2e) : µe = y and σ2
e =1n
n∑i=1
(yi − y)2 (Under Ω)
7
![Page 15: Math 362: Mathmatical Statistics IImath.emory.edu/~lchen41/teaching/2020_Spring/Chapter-7_full.pdf · Gosset worked at the Guinness Brewery in Dublin, Ireland, and was interested](https://reader036.vdocuments.site/reader036/viewer/2022071005/5fc1e109d75c9e182e3dd73a/html5/thumbnails/15.jpg)
Prob. 1 Find a test statistic for H0 : µ = µ0 v.s. H1 : µ 6= µ0, with σ2 unknown
Sol. Composite-vs-composite test with:
ω =
(µ, σ2) : µ = µ0, σ2 > 0
Ω =
(µ, σ2) : µ ∈ R, σ2 > 0
The MLE under the two spaces are:
ωe = (µe, σ2e) : µe = µ0 and σ2
e =1n
n∑i=1
(yi − µ0)2 (Under ω)
Ωe = (µe, σ2e) : µe = y and σ2
e =1n
n∑i=1
(yi − y)2 (Under Ω)
7
![Page 16: Math 362: Mathmatical Statistics IImath.emory.edu/~lchen41/teaching/2020_Spring/Chapter-7_full.pdf · Gosset worked at the Guinness Brewery in Dublin, Ireland, and was interested](https://reader036.vdocuments.site/reader036/viewer/2022071005/5fc1e109d75c9e182e3dd73a/html5/thumbnails/16.jpg)
L(µ, σ2) = (2πσ2)−n exp
(−1
2
n∑i=1
(yi − µσ
)2)
L(ωe) = · · · =
[ne−1
2π∑n
i=1(yi − µ0)2
]n/2
L(Ωe) = · · · =
[ne−1
2π∑n
i=1(yi − y)2
]n/2
8
![Page 17: Math 362: Mathmatical Statistics IImath.emory.edu/~lchen41/teaching/2020_Spring/Chapter-7_full.pdf · Gosset worked at the Guinness Brewery in Dublin, Ireland, and was interested](https://reader036.vdocuments.site/reader036/viewer/2022071005/5fc1e109d75c9e182e3dd73a/html5/thumbnails/17.jpg)
L(µ, σ2) = (2πσ2)−n exp
(−1
2
n∑i=1
(yi − µσ
)2)
L(ωe) = · · · =
[ne−1
2π∑n
i=1(yi − µ0)2
]n/2
L(Ωe) = · · · =
[ne−1
2π∑n
i=1(yi − y)2
]n/2
8
![Page 18: Math 362: Mathmatical Statistics IImath.emory.edu/~lchen41/teaching/2020_Spring/Chapter-7_full.pdf · Gosset worked at the Guinness Brewery in Dublin, Ireland, and was interested](https://reader036.vdocuments.site/reader036/viewer/2022071005/5fc1e109d75c9e182e3dd73a/html5/thumbnails/18.jpg)
Hence
λ =L(ωe)
L(Ωe)=
[ ∑ni=1(yi − y)2∑n
i=1(yi − µ0)2
]n/2
= · · · =
[1 +
n(y − µ0)2∑ni=1(yi − y)2
]−n/2
=
1 +1
n − 1
y − µ0√1
n−1
∑ni=1(yi − y)2
/√n
2−n/2
=
[1 +
1n − 1
(y − µ0
s /√
n
)2]−n/2
=
[1 +
t2
n − 1
]−n/2
, t =y − µ0
s /√
n
9
![Page 19: Math 362: Mathmatical Statistics IImath.emory.edu/~lchen41/teaching/2020_Spring/Chapter-7_full.pdf · Gosset worked at the Guinness Brewery in Dublin, Ireland, and was interested](https://reader036.vdocuments.site/reader036/viewer/2022071005/5fc1e109d75c9e182e3dd73a/html5/thumbnails/19.jpg)
Finally, the test statistic is
T =Y − µ0
S/√
n
with Y =1n
n∑i=1
Yi and S2 =1
n − 1
n∑i=1
(Yi − Y
)2.
Question: Find the exact distribution of T .
10
![Page 20: Math 362: Mathmatical Statistics IImath.emory.edu/~lchen41/teaching/2020_Spring/Chapter-7_full.pdf · Gosset worked at the Guinness Brewery in Dublin, Ireland, and was interested](https://reader036.vdocuments.site/reader036/viewer/2022071005/5fc1e109d75c9e182e3dd73a/html5/thumbnails/20.jpg)
Finally, the test statistic is
T =Y − µ0
S/√
n
with Y =1n
n∑i=1
Yi and S2 =1
n − 1
n∑i=1
(Yi − Y
)2.
Question: Find the exact distribution of T .
10
![Page 21: Math 362: Mathmatical Statistics IImath.emory.edu/~lchen41/teaching/2020_Spring/Chapter-7_full.pdf · Gosset worked at the Guinness Brewery in Dublin, Ireland, and was interested](https://reader036.vdocuments.site/reader036/viewer/2022071005/5fc1e109d75c9e182e3dd73a/html5/thumbnails/21.jpg)
Prob. 2 Find a test statistic for H0 : σ2 = σ20 v.s. H1 : σ2 6= σ2
0 , with µ unknown
Sol. Composite-vs-composite test with:
ω =
(µ, σ2) : µ ∈ R, σ2 = σ20
Ω =
(µ, σ2) : µ ∈ R, σ2 > 0
The MLE under the two spaces are:
ωe = (µe, σ2e) : µe = y and σ2
e = σ20 (Under ω)
Ωe = (µe, σ2e) : µe = y and σ2
e =1n
n∑i=1
(yi − y)2 (Under Ω)
11
![Page 22: Math 362: Mathmatical Statistics IImath.emory.edu/~lchen41/teaching/2020_Spring/Chapter-7_full.pdf · Gosset worked at the Guinness Brewery in Dublin, Ireland, and was interested](https://reader036.vdocuments.site/reader036/viewer/2022071005/5fc1e109d75c9e182e3dd73a/html5/thumbnails/22.jpg)
Prob. 2 Find a test statistic for H0 : σ2 = σ20 v.s. H1 : σ2 6= σ2
0 , with µ unknown
Sol. Composite-vs-composite test with:
ω =
(µ, σ2) : µ ∈ R, σ2 = σ20
Ω =
(µ, σ2) : µ ∈ R, σ2 > 0
The MLE under the two spaces are:
ωe = (µe, σ2e) : µe = y and σ2
e = σ20 (Under ω)
Ωe = (µe, σ2e) : µe = y and σ2
e =1n
n∑i=1
(yi − y)2 (Under Ω)
11
![Page 23: Math 362: Mathmatical Statistics IImath.emory.edu/~lchen41/teaching/2020_Spring/Chapter-7_full.pdf · Gosset worked at the Guinness Brewery in Dublin, Ireland, and was interested](https://reader036.vdocuments.site/reader036/viewer/2022071005/5fc1e109d75c9e182e3dd73a/html5/thumbnails/23.jpg)
Prob. 2 Find a test statistic for H0 : σ2 = σ20 v.s. H1 : σ2 6= σ2
0 , with µ unknown
Sol. Composite-vs-composite test with:
ω =
(µ, σ2) : µ ∈ R, σ2 = σ20
Ω =
(µ, σ2) : µ ∈ R, σ2 > 0
The MLE under the two spaces are:
ωe = (µe, σ2e) : µe = y and σ2
e = σ20 (Under ω)
Ωe = (µe, σ2e) : µe = y and σ2
e =1n
n∑i=1
(yi − y)2 (Under Ω)
11
![Page 24: Math 362: Mathmatical Statistics IImath.emory.edu/~lchen41/teaching/2020_Spring/Chapter-7_full.pdf · Gosset worked at the Guinness Brewery in Dublin, Ireland, and was interested](https://reader036.vdocuments.site/reader036/viewer/2022071005/5fc1e109d75c9e182e3dd73a/html5/thumbnails/24.jpg)
L(µ, σ2) = (2πσ2)−n exp
(−1
2
n∑i=1
(yi − µσ
)2)
L(ωe) = (2πσ2)−n exp
(−1
2
n∑i=1
(yi − yσ0
)2)
L(Ωe) = · · · =
[ne−1
2π∑n
i=1(yi − y)2
]n/2
12
![Page 25: Math 362: Mathmatical Statistics IImath.emory.edu/~lchen41/teaching/2020_Spring/Chapter-7_full.pdf · Gosset worked at the Guinness Brewery in Dublin, Ireland, and was interested](https://reader036.vdocuments.site/reader036/viewer/2022071005/5fc1e109d75c9e182e3dd73a/html5/thumbnails/25.jpg)
L(µ, σ2) = (2πσ2)−n exp
(−1
2
n∑i=1
(yi − µσ
)2)
L(ωe) = (2πσ2)−n exp
(−1
2
n∑i=1
(yi − yσ0
)2)
L(Ωe) = · · · =
[ne−1
2π∑n
i=1(yi − y)2
]n/2
12
![Page 26: Math 362: Mathmatical Statistics IImath.emory.edu/~lchen41/teaching/2020_Spring/Chapter-7_full.pdf · Gosset worked at the Guinness Brewery in Dublin, Ireland, and was interested](https://reader036.vdocuments.site/reader036/viewer/2022071005/5fc1e109d75c9e182e3dd73a/html5/thumbnails/26.jpg)
Hence
λ =L(ωe)
L(Ωe)=
[∑ni=1(yi − y)2
nσ20
]n/2
exp
(−1
2
n∑i=1
(yi − yσ0
)2
+n2
)
=
[1
n−1
∑ni=1(yi − y)2
nn−1σ
20
]n/2
exp
(−n − 1
2σ20
1n − 1
n∑i=1
(yi − y)2 +n2
)
=
[s2
nn−1σ
20
]n/2
exp
(−n − 1
2σ20
s2 +n2
)
13
![Page 27: Math 362: Mathmatical Statistics IImath.emory.edu/~lchen41/teaching/2020_Spring/Chapter-7_full.pdf · Gosset worked at the Guinness Brewery in Dublin, Ireland, and was interested](https://reader036.vdocuments.site/reader036/viewer/2022071005/5fc1e109d75c9e182e3dd73a/html5/thumbnails/27.jpg)
Finally, the test statistic is
S2 =1
n − 1
n∑i=1
(Yi − Y
)2with Y =
1n
n∑i=1
Yi
Question: Find the exact distribution of S2.
14
![Page 28: Math 362: Mathmatical Statistics IImath.emory.edu/~lchen41/teaching/2020_Spring/Chapter-7_full.pdf · Gosset worked at the Guinness Brewery in Dublin, Ireland, and was interested](https://reader036.vdocuments.site/reader036/viewer/2022071005/5fc1e109d75c9e182e3dd73a/html5/thumbnails/28.jpg)
Finally, the test statistic is
S2 =1
n − 1
n∑i=1
(Yi − Y
)2with Y =
1n
n∑i=1
Yi
Question: Find the exact distribution of S2.
14
![Page 29: Math 362: Mathmatical Statistics IImath.emory.edu/~lchen41/teaching/2020_Spring/Chapter-7_full.pdf · Gosset worked at the Guinness Brewery in Dublin, Ireland, and was interested](https://reader036.vdocuments.site/reader036/viewer/2022071005/5fc1e109d75c9e182e3dd73a/html5/thumbnails/29.jpg)
Plan
§ 7.1 Introduction
§ 7.2 Comparing Y−µσ/√
n and Y−µS/√
n
§ 7.3 Deriving the Distribution of Y−µS/√
n
§ 7.4 Drawing Inferences About µ
§ 7.5 Drawing Inferences About σ2
15
![Page 30: Math 362: Mathmatical Statistics IImath.emory.edu/~lchen41/teaching/2020_Spring/Chapter-7_full.pdf · Gosset worked at the Guinness Brewery in Dublin, Ireland, and was interested](https://reader036.vdocuments.site/reader036/viewer/2022071005/5fc1e109d75c9e182e3dd73a/html5/thumbnails/30.jpg)
Chapter 7. Inference Based on The Normal Distribution
§ 7.1 Introduction
§ 7.2 Comparing Y−µσ/√
n and Y−µS/√
n
§ 7.3 Deriving the Distribution of Y−µS/√
n
§ 7.4 Drawing Inferences About µ
§ 7.5 Drawing Inferences About σ2
16
![Page 31: Math 362: Mathmatical Statistics IImath.emory.edu/~lchen41/teaching/2020_Spring/Chapter-7_full.pdf · Gosset worked at the Guinness Brewery in Dublin, Ireland, and was interested](https://reader036.vdocuments.site/reader036/viewer/2022071005/5fc1e109d75c9e182e3dd73a/html5/thumbnails/31.jpg)
§ 7.2 Comparing Y−µσ/√
n and Y−µS/√
n
17
![Page 32: Math 362: Mathmatical Statistics IImath.emory.edu/~lchen41/teaching/2020_Spring/Chapter-7_full.pdf · Gosset worked at the Guinness Brewery in Dublin, Ireland, and was interested](https://reader036.vdocuments.site/reader036/viewer/2022071005/5fc1e109d75c9e182e3dd73a/html5/thumbnails/32.jpg)
Ref. Student’s t distribution comes from William Sealy Gosset’s 1908 paper in Biometrika underthe pseudonym "Student".
Gosset worked at the Guinness Brewery in Dublin, Ireland, and was interested in theproblems of small samples – for example, the chemical properties of barley where samplesizes might be as few as 3.
V1 One version of the origin of the pseudonym is that Gosset’s employer preferred staff to usepen names when publishing scientific papers instead of their real name, so he used the name"Student" to hide his identity.
V2 Another version is that Guinness did not want their competitors to know that they were usingthe t-test to determine the quality of raw material
18
![Page 33: Math 362: Mathmatical Statistics IImath.emory.edu/~lchen41/teaching/2020_Spring/Chapter-7_full.pdf · Gosset worked at the Guinness Brewery in Dublin, Ireland, and was interested](https://reader036.vdocuments.site/reader036/viewer/2022071005/5fc1e109d75c9e182e3dd73a/html5/thumbnails/33.jpg)
Ref. Student’s t distribution comes from William Sealy Gosset’s 1908 paper in Biometrika underthe pseudonym "Student".
Gosset worked at the Guinness Brewery in Dublin, Ireland, and was interested in theproblems of small samples – for example, the chemical properties of barley where samplesizes might be as few as 3.
V1 One version of the origin of the pseudonym is that Gosset’s employer preferred staff to usepen names when publishing scientific papers instead of their real name, so he used the name"Student" to hide his identity.
V2 Another version is that Guinness did not want their competitors to know that they were usingthe t-test to determine the quality of raw material
18
![Page 34: Math 362: Mathmatical Statistics IImath.emory.edu/~lchen41/teaching/2020_Spring/Chapter-7_full.pdf · Gosset worked at the Guinness Brewery in Dublin, Ireland, and was interested](https://reader036.vdocuments.site/reader036/viewer/2022071005/5fc1e109d75c9e182e3dd73a/html5/thumbnails/34.jpg)
Ref. Student’s t distribution comes from William Sealy Gosset’s 1908 paper in Biometrika underthe pseudonym "Student".
Gosset worked at the Guinness Brewery in Dublin, Ireland, and was interested in theproblems of small samples – for example, the chemical properties of barley where samplesizes might be as few as 3.
V1 One version of the origin of the pseudonym is that Gosset’s employer preferred staff to usepen names when publishing scientific papers instead of their real name, so he used the name"Student" to hide his identity.
V2 Another version is that Guinness did not want their competitors to know that they were usingthe t-test to determine the quality of raw material
18
![Page 35: Math 362: Mathmatical Statistics IImath.emory.edu/~lchen41/teaching/2020_Spring/Chapter-7_full.pdf · Gosset worked at the Guinness Brewery in Dublin, Ireland, and was interested](https://reader036.vdocuments.site/reader036/viewer/2022071005/5fc1e109d75c9e182e3dd73a/html5/thumbnails/35.jpg)
Ref. Student’s t distribution comes from William Sealy Gosset’s 1908 paper in Biometrika underthe pseudonym "Student".
Gosset worked at the Guinness Brewery in Dublin, Ireland, and was interested in theproblems of small samples – for example, the chemical properties of barley where samplesizes might be as few as 3.
V1 One version of the origin of the pseudonym is that Gosset’s employer preferred staff to usepen names when publishing scientific papers instead of their real name, so he used the name"Student" to hide his identity.
V2 Another version is that Guinness did not want their competitors to know that they were usingthe t-test to determine the quality of raw material
18
![Page 36: Math 362: Mathmatical Statistics IImath.emory.edu/~lchen41/teaching/2020_Spring/Chapter-7_full.pdf · Gosset worked at the Guinness Brewery in Dublin, Ireland, and was interested](https://reader036.vdocuments.site/reader036/viewer/2022071005/5fc1e109d75c9e182e3dd73a/html5/thumbnails/36.jpg)
Plan
§ 7.1 Introduction
§ 7.2 Comparing Y−µσ/√
n and Y−µS/√
n
§ 7.3 Deriving the Distribution of Y−µS/√
n
§ 7.4 Drawing Inferences About µ
§ 7.5 Drawing Inferences About σ2
19
![Page 37: Math 362: Mathmatical Statistics IImath.emory.edu/~lchen41/teaching/2020_Spring/Chapter-7_full.pdf · Gosset worked at the Guinness Brewery in Dublin, Ireland, and was interested](https://reader036.vdocuments.site/reader036/viewer/2022071005/5fc1e109d75c9e182e3dd73a/html5/thumbnails/37.jpg)
Chapter 7. Inference Based on The Normal Distribution
§ 7.1 Introduction
§ 7.2 Comparing Y−µσ/√
n and Y−µS/√
n
§ 7.3 Deriving the Distribution of Y−µS/√
n
§ 7.4 Drawing Inferences About µ
§ 7.5 Drawing Inferences About σ2
20
![Page 38: Math 362: Mathmatical Statistics IImath.emory.edu/~lchen41/teaching/2020_Spring/Chapter-7_full.pdf · Gosset worked at the Guinness Brewery in Dublin, Ireland, and was interested](https://reader036.vdocuments.site/reader036/viewer/2022071005/5fc1e109d75c9e182e3dd73a/html5/thumbnails/38.jpg)
§ 7.3 Deriving the Distribution of Y−µS/√
n
Def. Sampling distributions
Distributions of functions of random sample of given size.statistics / estimators
E.g. A random sample of size n from N(µ, σ2) with σ2 known.
Sample mean Y = 1n
∑ni=1 Yi ∼ N(µ, σ2/n)
Aim: Determine distributions for
Sample variance S2 := 1n−1
∑ni=1
(Yi − Y
)2Chi square distr.
T :=Y − µS/√
nStudent t distr.
S21
σ21
/S2
1
σ21
F distr.
21
![Page 39: Math 362: Mathmatical Statistics IImath.emory.edu/~lchen41/teaching/2020_Spring/Chapter-7_full.pdf · Gosset worked at the Guinness Brewery in Dublin, Ireland, and was interested](https://reader036.vdocuments.site/reader036/viewer/2022071005/5fc1e109d75c9e182e3dd73a/html5/thumbnails/39.jpg)
§ 7.3 Deriving the Distribution of Y−µS/√
n
Def. Sampling distributions
Distributions of functions of random sample of given size.statistics / estimators
E.g. A random sample of size n from N(µ, σ2) with σ2 known.
Sample mean Y = 1n
∑ni=1 Yi ∼ N(µ, σ2/n)
Aim: Determine distributions for
Sample variance S2 := 1n−1
∑ni=1
(Yi − Y
)2Chi square distr.
T :=Y − µS/√
nStudent t distr.
S21
σ21
/S2
1
σ21
F distr.
21
![Page 40: Math 362: Mathmatical Statistics IImath.emory.edu/~lchen41/teaching/2020_Spring/Chapter-7_full.pdf · Gosset worked at the Guinness Brewery in Dublin, Ireland, and was interested](https://reader036.vdocuments.site/reader036/viewer/2022071005/5fc1e109d75c9e182e3dd73a/html5/thumbnails/40.jpg)
§ 7.3 Deriving the Distribution of Y−µS/√
n
Def. Sampling distributions
Distributions of functions of random sample of given size.statistics / estimators
E.g. A random sample of size n from N(µ, σ2) with σ2 known.
Sample mean Y = 1n
∑ni=1 Yi ∼ N(µ, σ2/n)
Aim: Determine distributions for
Sample variance S2 := 1n−1
∑ni=1
(Yi − Y
)2Chi square distr.
T :=Y − µS/√
nStudent t distr.
S21
σ21
/S2
1
σ21
F distr.
21
![Page 41: Math 362: Mathmatical Statistics IImath.emory.edu/~lchen41/teaching/2020_Spring/Chapter-7_full.pdf · Gosset worked at the Guinness Brewery in Dublin, Ireland, and was interested](https://reader036.vdocuments.site/reader036/viewer/2022071005/5fc1e109d75c9e182e3dd73a/html5/thumbnails/41.jpg)
§ 7.3 Deriving the Distribution of Y−µS/√
n
Def. Sampling distributions
Distributions of functions of random sample of given size.statistics / estimators
E.g. A random sample of size n from N(µ, σ2) with σ2 known.
Sample mean Y = 1n
∑ni=1 Yi ∼ N(µ, σ2/n)
Aim: Determine distributions for
Sample variance S2 := 1n−1
∑ni=1
(Yi − Y
)2Chi square distr.
T :=Y − µS/√
nStudent t distr.
S21
σ21
/S2
1
σ21
F distr.
21
![Page 42: Math 362: Mathmatical Statistics IImath.emory.edu/~lchen41/teaching/2020_Spring/Chapter-7_full.pdf · Gosset worked at the Guinness Brewery in Dublin, Ireland, and was interested](https://reader036.vdocuments.site/reader036/viewer/2022071005/5fc1e109d75c9e182e3dd73a/html5/thumbnails/42.jpg)
§ 7.3 Deriving the Distribution of Y−µS/√
n
Def. Sampling distributions
Distributions of functions of random sample of given size.statistics / estimators
E.g. A random sample of size n from N(µ, σ2) with σ2 known.
Sample mean Y = 1n
∑ni=1 Yi ∼ N(µ, σ2/n)
Aim: Determine distributions for
Sample variance S2 := 1n−1
∑ni=1
(Yi − Y
)2Chi square distr.
T :=Y − µS/√
nStudent t distr.
S21
σ21
/S2
1
σ21
F distr.
21
![Page 43: Math 362: Mathmatical Statistics IImath.emory.edu/~lchen41/teaching/2020_Spring/Chapter-7_full.pdf · Gosset worked at the Guinness Brewery in Dublin, Ireland, and was interested](https://reader036.vdocuments.site/reader036/viewer/2022071005/5fc1e109d75c9e182e3dd73a/html5/thumbnails/43.jpg)
§ 7.3 Deriving the Distribution of Y−µS/√
n
Def. Sampling distributions
Distributions of functions of random sample of given size.statistics / estimators
E.g. A random sample of size n from N(µ, σ2) with σ2 known.
Sample mean Y = 1n
∑ni=1 Yi ∼ N(µ, σ2/n)
Aim: Determine distributions for
Sample variance S2 := 1n−1
∑ni=1
(Yi − Y
)2Chi square distr.
T :=Y − µS/√
nStudent t distr.
S21
σ21
/S2
1
σ21
F distr.
21
![Page 44: Math 362: Mathmatical Statistics IImath.emory.edu/~lchen41/teaching/2020_Spring/Chapter-7_full.pdf · Gosset worked at the Guinness Brewery in Dublin, Ireland, and was interested](https://reader036.vdocuments.site/reader036/viewer/2022071005/5fc1e109d75c9e182e3dd73a/html5/thumbnails/44.jpg)
Thm 1. Let U =∑m
i=1 Z 2j , where Zj are independent N(0, 1) normal r.v.s. Then
U ∼ Gamma(shape=m/2,rate=1/2).
namely,
fU(u) =1
2m/2Γ(m/2)u
m2 −1e−u/2, u ≥ 0
Def 1. U in Thm 1 is called chi square distribution with m dgs of freedom.
22
![Page 45: Math 362: Mathmatical Statistics IImath.emory.edu/~lchen41/teaching/2020_Spring/Chapter-7_full.pdf · Gosset worked at the Guinness Brewery in Dublin, Ireland, and was interested](https://reader036.vdocuments.site/reader036/viewer/2022071005/5fc1e109d75c9e182e3dd73a/html5/thumbnails/45.jpg)
Thm 1. Let U =∑m
i=1 Z 2j , where Zj are independent N(0, 1) normal r.v.s. Then
U ∼ Gamma(shape=m/2,rate=1/2).
namely,
fU(u) =1
2m/2Γ(m/2)u
m2 −1e−u/2, u ≥ 0
Def 1. U in Thm 1 is called chi square distribution with m dgs of freedom.
22
![Page 46: Math 362: Mathmatical Statistics IImath.emory.edu/~lchen41/teaching/2020_Spring/Chapter-7_full.pdf · Gosset worked at the Guinness Brewery in Dublin, Ireland, and was interested](https://reader036.vdocuments.site/reader036/viewer/2022071005/5fc1e109d75c9e182e3dd73a/html5/thumbnails/46.jpg)
23
![Page 47: Math 362: Mathmatical Statistics IImath.emory.edu/~lchen41/teaching/2020_Spring/Chapter-7_full.pdf · Gosset worked at the Guinness Brewery in Dublin, Ireland, and was interested](https://reader036.vdocuments.site/reader036/viewer/2022071005/5fc1e109d75c9e182e3dd73a/html5/thumbnails/47.jpg)
Chi Square Table
P(χ25 ≤ 1.145) = 0.05 ⇐⇒ χ2
0.05,5 = 1.145
P(χ25 ≤ 15.086) = 0.99 ⇐⇒ χ2
0.99,5 = 15.086
1 > pchisq(1.145, df = 5)2 [1] 0.049956223 > pchisq(15.086, df = 5)4 [1] 0.9899989
1 > qchisq(0.05, df = 5)2 [1] 1.1454763 > qchisq(0.99, df = 5)4 [1] 15.08627
24
![Page 48: Math 362: Mathmatical Statistics IImath.emory.edu/~lchen41/teaching/2020_Spring/Chapter-7_full.pdf · Gosset worked at the Guinness Brewery in Dublin, Ireland, and was interested](https://reader036.vdocuments.site/reader036/viewer/2022071005/5fc1e109d75c9e182e3dd73a/html5/thumbnails/48.jpg)
Thm 2. Let Y1, · · · ,Yn be a random sample from N(µ, σ2). Then
(a) S2 and Y are independent.
(b)(n − 1)S2
σ2 =1σ2
n∑i=1
(Yi − Y
)2∼ Chi Square(n − 1).
Proof. We will prove the case n = 2.
Y =Y1 + Y2
2, Y1 − Y =
Y1 − Y2
2, Y2 − Y =
Y2 − Y1
2
S2 = ... =12
(Y1 − Y2)2
(a) It is equivalanet to show Y1 + Y2 ⊥ Y1 − Y2. Since they are normal, itsuffices to show that
E[(Y1 + Y2)(Y1 − Y2)] = E[Y1 + Y2]E[Y1 − Y2]
(b) (n−1)S2
σ2 =(
Y1−Y2√2σ
)2and Y1−Y2√
2σ∼ N(0, 1) ...
25
![Page 49: Math 362: Mathmatical Statistics IImath.emory.edu/~lchen41/teaching/2020_Spring/Chapter-7_full.pdf · Gosset worked at the Guinness Brewery in Dublin, Ireland, and was interested](https://reader036.vdocuments.site/reader036/viewer/2022071005/5fc1e109d75c9e182e3dd73a/html5/thumbnails/49.jpg)
Thm 2. Let Y1, · · · ,Yn be a random sample from N(µ, σ2). Then
(a) S2 and Y are independent.
(b)(n − 1)S2
σ2 =1σ2
n∑i=1
(Yi − Y
)2∼ Chi Square(n − 1).
Proof. We will prove the case n = 2.
Y =Y1 + Y2
2, Y1 − Y =
Y1 − Y2
2, Y2 − Y =
Y2 − Y1
2
S2 = ... =12
(Y1 − Y2)2
(a) It is equivalanet to show Y1 + Y2 ⊥ Y1 − Y2. Since they are normal, itsuffices to show that
E[(Y1 + Y2)(Y1 − Y2)] = E[Y1 + Y2]E[Y1 − Y2]
(b) (n−1)S2
σ2 =(
Y1−Y2√2σ
)2and Y1−Y2√
2σ∼ N(0, 1) ...
25
![Page 50: Math 362: Mathmatical Statistics IImath.emory.edu/~lchen41/teaching/2020_Spring/Chapter-7_full.pdf · Gosset worked at the Guinness Brewery in Dublin, Ireland, and was interested](https://reader036.vdocuments.site/reader036/viewer/2022071005/5fc1e109d75c9e182e3dd73a/html5/thumbnails/50.jpg)
Thm 2. Let Y1, · · · ,Yn be a random sample from N(µ, σ2). Then
(a) S2 and Y are independent.
(b)(n − 1)S2
σ2 =1σ2
n∑i=1
(Yi − Y
)2∼ Chi Square(n − 1).
Proof. We will prove the case n = 2.
Y =Y1 + Y2
2, Y1 − Y =
Y1 − Y2
2, Y2 − Y =
Y2 − Y1
2
S2 = ... =12
(Y1 − Y2)2
(a) It is equivalanet to show Y1 + Y2 ⊥ Y1 − Y2. Since they are normal, itsuffices to show that
E[(Y1 + Y2)(Y1 − Y2)] = E[Y1 + Y2]E[Y1 − Y2]
(b) (n−1)S2
σ2 =(
Y1−Y2√2σ
)2and Y1−Y2√
2σ∼ N(0, 1) ...
25
![Page 51: Math 362: Mathmatical Statistics IImath.emory.edu/~lchen41/teaching/2020_Spring/Chapter-7_full.pdf · Gosset worked at the Guinness Brewery in Dublin, Ireland, and was interested](https://reader036.vdocuments.site/reader036/viewer/2022071005/5fc1e109d75c9e182e3dd73a/html5/thumbnails/51.jpg)
Thm 2. Let Y1, · · · ,Yn be a random sample from N(µ, σ2). Then
(a) S2 and Y are independent.
(b)(n − 1)S2
σ2 =1σ2
n∑i=1
(Yi − Y
)2∼ Chi Square(n − 1).
Proof. We will prove the case n = 2.
Y =Y1 + Y2
2, Y1 − Y =
Y1 − Y2
2, Y2 − Y =
Y2 − Y1
2
S2 = ... =12
(Y1 − Y2)2
(a) It is equivalanet to show Y1 + Y2 ⊥ Y1 − Y2. Since they are normal, itsuffices to show that
E[(Y1 + Y2)(Y1 − Y2)] = E[Y1 + Y2]E[Y1 − Y2]
(b) (n−1)S2
σ2 =(
Y1−Y2√2σ
)2and Y1−Y2√
2σ∼ N(0, 1) ...
25
![Page 52: Math 362: Mathmatical Statistics IImath.emory.edu/~lchen41/teaching/2020_Spring/Chapter-7_full.pdf · Gosset worked at the Guinness Brewery in Dublin, Ireland, and was interested](https://reader036.vdocuments.site/reader036/viewer/2022071005/5fc1e109d75c9e182e3dd73a/html5/thumbnails/52.jpg)
Thm 2. Let Y1, · · · ,Yn be a random sample from N(µ, σ2). Then
(a) S2 and Y are independent.
(b)(n − 1)S2
σ2 =1σ2
n∑i=1
(Yi − Y
)2∼ Chi Square(n − 1).
Proof. We will prove the case n = 2.
Y =Y1 + Y2
2, Y1 − Y =
Y1 − Y2
2, Y2 − Y =
Y2 − Y1
2
S2 = ... =12
(Y1 − Y2)2
(a) It is equivalanet to show Y1 + Y2 ⊥ Y1 − Y2. Since they are normal, itsuffices to show that
E[(Y1 + Y2)(Y1 − Y2)] = E[Y1 + Y2]E[Y1 − Y2]
(b) (n−1)S2
σ2 =(
Y1−Y2√2σ
)2and Y1−Y2√
2σ∼ N(0, 1) ...
25
![Page 53: Math 362: Mathmatical Statistics IImath.emory.edu/~lchen41/teaching/2020_Spring/Chapter-7_full.pdf · Gosset worked at the Guinness Brewery in Dublin, Ireland, and was interested](https://reader036.vdocuments.site/reader036/viewer/2022071005/5fc1e109d75c9e182e3dd73a/html5/thumbnails/53.jpg)
Thm 2. Let Y1, · · · ,Yn be a random sample from N(µ, σ2). Then
(a) S2 and Y are independent.
(b)(n − 1)S2
σ2 =1σ2
n∑i=1
(Yi − Y
)2∼ Chi Square(n − 1).
Proof. We will prove the case n = 2.
Y =Y1 + Y2
2, Y1 − Y =
Y1 − Y2
2, Y2 − Y =
Y2 − Y1
2
S2 = ... =12
(Y1 − Y2)2
(a) It is equivalanet to show Y1 + Y2 ⊥ Y1 − Y2. Since they are normal, itsuffices to show that
E[(Y1 + Y2)(Y1 − Y2)] = E[Y1 + Y2]E[Y1 − Y2]
(b) (n−1)S2
σ2 =(
Y1−Y2√2σ
)2and Y1−Y2√
2σ∼ N(0, 1) ...
25
![Page 54: Math 362: Mathmatical Statistics IImath.emory.edu/~lchen41/teaching/2020_Spring/Chapter-7_full.pdf · Gosset worked at the Guinness Brewery in Dublin, Ireland, and was interested](https://reader036.vdocuments.site/reader036/viewer/2022071005/5fc1e109d75c9e182e3dd73a/html5/thumbnails/54.jpg)
Def 2. If U ∼ Chi Square(n) and V ∼ Chi Square(m), and U ⊥ V , then
F :=V/mU/n
follows the (Snedecor’s) F distribution with m and n degrees offreedom.
Thm 3. Let Fm,n = V/mU/n be an F r.v. with m and n degrees of freedom. Then
fFm,n (w) =Γ(m+n
2
)mm/2nn/2
Γ(m/2)Γ(n/2)× wm/2−1
(n + mw)(m+n)/2 , w ≥ 0
Equivalently,
fFm,n (w) = B(m/2, n/2)−1(m
n
)m2
wm2 −1
(1 +
mn
w)−m+n
2
where B(a, b) = Γ(a)Γ(b)/Γ(a + b)
26
![Page 55: Math 362: Mathmatical Statistics IImath.emory.edu/~lchen41/teaching/2020_Spring/Chapter-7_full.pdf · Gosset worked at the Guinness Brewery in Dublin, Ireland, and was interested](https://reader036.vdocuments.site/reader036/viewer/2022071005/5fc1e109d75c9e182e3dd73a/html5/thumbnails/55.jpg)
Def 2. If U ∼ Chi Square(n) and V ∼ Chi Square(m), and U ⊥ V , then
F :=V/mU/n
follows the (Snedecor’s) F distribution with m and n degrees offreedom.
Thm 3. Let Fm,n = V/mU/n be an F r.v. with m and n degrees of freedom. Then
fFm,n (w) =Γ(m+n
2
)mm/2nn/2
Γ(m/2)Γ(n/2)× wm/2−1
(n + mw)(m+n)/2 , w ≥ 0
Equivalently,
fFm,n (w) = B(m/2, n/2)−1(m
n
)m2
wm2 −1
(1 +
mn
w)−m+n
2
where B(a, b) = Γ(a)Γ(b)/Γ(a + b)
26
![Page 56: Math 362: Mathmatical Statistics IImath.emory.edu/~lchen41/teaching/2020_Spring/Chapter-7_full.pdf · Gosset worked at the Guinness Brewery in Dublin, Ireland, and was interested](https://reader036.vdocuments.site/reader036/viewer/2022071005/5fc1e109d75c9e182e3dd73a/html5/thumbnails/56.jpg)
Def 2. If U ∼ Chi Square(n) and V ∼ Chi Square(m), and U ⊥ V , then
F :=V/mU/n
follows the (Snedecor’s) F distribution with m and n degrees offreedom.
Thm 3. Let Fm,n = V/mU/n be an F r.v. with m and n degrees of freedom. Then
fFm,n (w) =Γ(m+n
2
)mm/2nn/2
Γ(m/2)Γ(n/2)× wm/2−1
(n + mw)(m+n)/2 , w ≥ 0
Equivalently,
fFm,n (w) = B(m/2, n/2)−1(m
n
)m2
wm2 −1
(1 +
mn
w)−m+n
2
where B(a, b) = Γ(a)Γ(b)/Γ(a + b)
26
![Page 57: Math 362: Mathmatical Statistics IImath.emory.edu/~lchen41/teaching/2020_Spring/Chapter-7_full.pdf · Gosset worked at the Guinness Brewery in Dublin, Ireland, and was interested](https://reader036.vdocuments.site/reader036/viewer/2022071005/5fc1e109d75c9e182e3dd73a/html5/thumbnails/57.jpg)
Def 2. If U ∼ Chi Square(n) and V ∼ Chi Square(m), and U ⊥ V , then
F :=V/mU/n
follows the (Snedecor’s) F distribution with m and n degrees offreedom.
Thm 3. Let Fm,n = V/mU/n be an F r.v. with m and n degrees of freedom. Then
fFm,n (w) =Γ(m+n
2
)mm/2nn/2
Γ(m/2)Γ(n/2)× wm/2−1
(n + mw)(m+n)/2 , w ≥ 0
Equivalently,
fFm,n (w) = B(m/2, n/2)−1(m
n
)m2
wm2 −1
(1 +
mn
w)−m+n
2
where B(a, b) = Γ(a)Γ(b)/Γ(a + b)
26
![Page 58: Math 362: Mathmatical Statistics IImath.emory.edu/~lchen41/teaching/2020_Spring/Chapter-7_full.pdf · Gosset worked at the Guinness Brewery in Dublin, Ireland, and was interested](https://reader036.vdocuments.site/reader036/viewer/2022071005/5fc1e109d75c9e182e3dd73a/html5/thumbnails/58.jpg)
Recall
Proof of Thm 3.
27
![Page 59: Math 362: Mathmatical Statistics IImath.emory.edu/~lchen41/teaching/2020_Spring/Chapter-7_full.pdf · Gosset worked at the Guinness Brewery in Dublin, Ireland, and was interested](https://reader036.vdocuments.site/reader036/viewer/2022071005/5fc1e109d75c9e182e3dd73a/html5/thumbnails/59.jpg)
Recall
Proof of Thm 3.
27
![Page 60: Math 362: Mathmatical Statistics IImath.emory.edu/~lchen41/teaching/2020_Spring/Chapter-7_full.pdf · Gosset worked at the Guinness Brewery in Dublin, Ireland, and was interested](https://reader036.vdocuments.site/reader036/viewer/2022071005/5fc1e109d75c9e182e3dd73a/html5/thumbnails/60.jpg)
Recall
Proof of Thm 3.
27
![Page 61: Math 362: Mathmatical Statistics IImath.emory.edu/~lchen41/teaching/2020_Spring/Chapter-7_full.pdf · Gosset worked at the Guinness Brewery in Dublin, Ireland, and was interested](https://reader036.vdocuments.site/reader036/viewer/2022071005/5fc1e109d75c9e182e3dd73a/html5/thumbnails/61.jpg)
28
![Page 62: Math 362: Mathmatical Statistics IImath.emory.edu/~lchen41/teaching/2020_Spring/Chapter-7_full.pdf · Gosset worked at the Guinness Brewery in Dublin, Ireland, and was interested](https://reader036.vdocuments.site/reader036/viewer/2022071005/5fc1e109d75c9e182e3dd73a/html5/thumbnails/62.jpg)
1 # Draw F density2 x=seq(0,5,0.01)3 pdf= cbind(df(x, df1 = 1, df2 = 1),4 df(x, df1 = 2, df2 = 1),5 df(x, df1 = 5, df2 = 2),6 df(x, df1 = 10, df2 = 1),7 df(x, df1 = 100, df2 = 100))8 matplot(x,pdf, type = " l " )9 title ("F with various dgrs of freedom")
29
![Page 63: Math 362: Mathmatical Statistics IImath.emory.edu/~lchen41/teaching/2020_Spring/Chapter-7_full.pdf · Gosset worked at the Guinness Brewery in Dublin, Ireland, and was interested](https://reader036.vdocuments.site/reader036/viewer/2022071005/5fc1e109d75c9e182e3dd73a/html5/thumbnails/63.jpg)
F Table
P(F3,5 ≤ 5.41) = 0.95 ⇐⇒ F0.95,3,5 = 5.41
1 > pf(5.41,df1 = 3, df2 = 5)2 [1] 0.9500093
1 > qf(0.95,df1 = 3, df2 = 5)2 [1] 5.409451
30
![Page 64: Math 362: Mathmatical Statistics IImath.emory.edu/~lchen41/teaching/2020_Spring/Chapter-7_full.pdf · Gosset worked at the Guinness Brewery in Dublin, Ireland, and was interested](https://reader036.vdocuments.site/reader036/viewer/2022071005/5fc1e109d75c9e182e3dd73a/html5/thumbnails/64.jpg)
Def 3. Suppose Z ∼ N(0, 1), U ∼ Chi Square(n), and Z ⊥ U. Then
Tn =Z√U/n
follows the Student’s t-distribution of n degrees of freedom.
Remark T 2n ∼ F -distribution with 1 and n degrees of freedom.
Thm 4. The pdf of the Student t of degree n is
fTn (t) =Γ( n+1
2
)√
nπΓ( n
2
) × (1 +t2
n
)− n+22
, t ∈ R.
31
![Page 65: Math 362: Mathmatical Statistics IImath.emory.edu/~lchen41/teaching/2020_Spring/Chapter-7_full.pdf · Gosset worked at the Guinness Brewery in Dublin, Ireland, and was interested](https://reader036.vdocuments.site/reader036/viewer/2022071005/5fc1e109d75c9e182e3dd73a/html5/thumbnails/65.jpg)
Def 3. Suppose Z ∼ N(0, 1), U ∼ Chi Square(n), and Z ⊥ U. Then
Tn =Z√U/n
follows the Student’s t-distribution of n degrees of freedom.
Remark T 2n ∼ F -distribution with 1 and n degrees of freedom.
Thm 4. The pdf of the Student t of degree n is
fTn (t) =Γ( n+1
2
)√
nπΓ( n
2
) × (1 +t2
n
)− n+22
, t ∈ R.
31
![Page 66: Math 362: Mathmatical Statistics IImath.emory.edu/~lchen41/teaching/2020_Spring/Chapter-7_full.pdf · Gosset worked at the Guinness Brewery in Dublin, Ireland, and was interested](https://reader036.vdocuments.site/reader036/viewer/2022071005/5fc1e109d75c9e182e3dd73a/html5/thumbnails/66.jpg)
Def 3. Suppose Z ∼ N(0, 1), U ∼ Chi Square(n), and Z ⊥ U. Then
Tn =Z√U/n
follows the Student’s t-distribution of n degrees of freedom.
Remark T 2n ∼ F -distribution with 1 and n degrees of freedom.
Thm 4. The pdf of the Student t of degree n is
fTn (t) =Γ( n+1
2
)√
nπΓ( n
2
) × (1 +t2
n
)− n+22
, t ∈ R.
31
![Page 67: Math 362: Mathmatical Statistics IImath.emory.edu/~lchen41/teaching/2020_Spring/Chapter-7_full.pdf · Gosset worked at the Guinness Brewery in Dublin, Ireland, and was interested](https://reader036.vdocuments.site/reader036/viewer/2022071005/5fc1e109d75c9e182e3dd73a/html5/thumbnails/67.jpg)
32
![Page 68: Math 362: Mathmatical Statistics IImath.emory.edu/~lchen41/teaching/2020_Spring/Chapter-7_full.pdf · Gosset worked at the Guinness Brewery in Dublin, Ireland, and was interested](https://reader036.vdocuments.site/reader036/viewer/2022071005/5fc1e109d75c9e182e3dd73a/html5/thumbnails/68.jpg)
1 # Draw Student t−density2 x=seq(−5,5,0.01)3 pdf= cbind(dt(x, df = 1),4 dt(x, df = 2),5 dt(x, df = 5),6 dt(x, df = 100))7 matplot(x,pdf, type = " l " )8 title ("Student’s t−distributions")
33
![Page 69: Math 362: Mathmatical Statistics IImath.emory.edu/~lchen41/teaching/2020_Spring/Chapter-7_full.pdf · Gosset worked at the Guinness Brewery in Dublin, Ireland, and was interested](https://reader036.vdocuments.site/reader036/viewer/2022071005/5fc1e109d75c9e182e3dd73a/html5/thumbnails/69.jpg)
t Table
P (T3 > 4.541) = 0.01 ⇐⇒ t0.01,3 = 4.541
1 > 1−pt(4.541, df =3)2 [1] 0.009998238
1 > alpha = 0.012 > qt(1−alpha, df = 3)3 [1] 4.540703
34
![Page 70: Math 362: Mathmatical Statistics IImath.emory.edu/~lchen41/teaching/2020_Spring/Chapter-7_full.pdf · Gosset worked at the Guinness Brewery in Dublin, Ireland, and was interested](https://reader036.vdocuments.site/reader036/viewer/2022071005/5fc1e109d75c9e182e3dd73a/html5/thumbnails/70.jpg)
Thm 5. Let Y1, · · · ,Yn be a random sample from N(µ, σ2). Then
Tn−1 =Y − µS/√
n∼ Student’s t of degree n − 1.
Proof.
Y − µS/√
n=
Y − µσ/√
n√(n − 1)S2
σ2(n − 1)
Y − µσ/√
n∼ N(0, 1) ⊥ (n − 1)S2
σ2 ∼ Chi Square(n − 1)
By Def. 2 ...
35
![Page 71: Math 362: Mathmatical Statistics IImath.emory.edu/~lchen41/teaching/2020_Spring/Chapter-7_full.pdf · Gosset worked at the Guinness Brewery in Dublin, Ireland, and was interested](https://reader036.vdocuments.site/reader036/viewer/2022071005/5fc1e109d75c9e182e3dd73a/html5/thumbnails/71.jpg)
Thm 5. Let Y1, · · · ,Yn be a random sample from N(µ, σ2). Then
Tn−1 =Y − µS/√
n∼ Student’s t of degree n − 1.
Proof.
Y − µS/√
n=
Y − µσ/√
n√(n − 1)S2
σ2(n − 1)
Y − µσ/√
n∼ N(0, 1) ⊥ (n − 1)S2
σ2 ∼ Chi Square(n − 1)
By Def. 2 ...
35
![Page 72: Math 362: Mathmatical Statistics IImath.emory.edu/~lchen41/teaching/2020_Spring/Chapter-7_full.pdf · Gosset worked at the Guinness Brewery in Dublin, Ireland, and was interested](https://reader036.vdocuments.site/reader036/viewer/2022071005/5fc1e109d75c9e182e3dd73a/html5/thumbnails/72.jpg)
Thm 5. Let Y1, · · · ,Yn be a random sample from N(µ, σ2). Then
Tn−1 =Y − µS/√
n∼ Student’s t of degree n − 1.
Proof.
Y − µS/√
n=
Y − µσ/√
n√(n − 1)S2
σ2(n − 1)
Y − µσ/√
n∼ N(0, 1) ⊥ (n − 1)S2
σ2 ∼ Chi Square(n − 1)
By Def. 2 ...
35
![Page 73: Math 362: Mathmatical Statistics IImath.emory.edu/~lchen41/teaching/2020_Spring/Chapter-7_full.pdf · Gosset worked at the Guinness Brewery in Dublin, Ireland, and was interested](https://reader036.vdocuments.site/reader036/viewer/2022071005/5fc1e109d75c9e182e3dd73a/html5/thumbnails/73.jpg)
Thm 5. Let Y1, · · · ,Yn be a random sample from N(µ, σ2). Then
Tn−1 =Y − µS/√
n∼ Student’s t of degree n − 1.
Proof.
Y − µS/√
n=
Y − µσ/√
n√(n − 1)S2
σ2(n − 1)
Y − µσ/√
n∼ N(0, 1) ⊥ (n − 1)S2
σ2 ∼ Chi Square(n − 1)
By Def. 2 ...
35
![Page 74: Math 362: Mathmatical Statistics IImath.emory.edu/~lchen41/teaching/2020_Spring/Chapter-7_full.pdf · Gosset worked at the Guinness Brewery in Dublin, Ireland, and was interested](https://reader036.vdocuments.site/reader036/viewer/2022071005/5fc1e109d75c9e182e3dd73a/html5/thumbnails/74.jpg)
As n→∞, Students’ t distribution will converge to N(0, 1):
Thm 6. fTn (x)→ fZ (x) =1√2π
e−x22 as n→∞, where Z ∼ N(0, 1).
Proof By Stirling’s formula:
Γ(z) =
√2πz
(ze
)z(1 + O(1/z)) as z →∞
=⇒ limn→∞
Γ( n+1
2
)√
nπ Γ( n
2
) =1√2π
......
36
![Page 75: Math 362: Mathmatical Statistics IImath.emory.edu/~lchen41/teaching/2020_Spring/Chapter-7_full.pdf · Gosset worked at the Guinness Brewery in Dublin, Ireland, and was interested](https://reader036.vdocuments.site/reader036/viewer/2022071005/5fc1e109d75c9e182e3dd73a/html5/thumbnails/75.jpg)
As n→∞, Students’ t distribution will converge to N(0, 1):
Thm 6. fTn (x)→ fZ (x) =1√2π
e−x22 as n→∞, where Z ∼ N(0, 1).
Proof By Stirling’s formula:
Γ(z) =
√2πz
(ze
)z(1 + O(1/z)) as z →∞
=⇒ limn→∞
Γ( n+1
2
)√
nπ Γ( n
2
) =1√2π
......
36
![Page 76: Math 362: Mathmatical Statistics IImath.emory.edu/~lchen41/teaching/2020_Spring/Chapter-7_full.pdf · Gosset worked at the Guinness Brewery in Dublin, Ireland, and was interested](https://reader036.vdocuments.site/reader036/viewer/2022071005/5fc1e109d75c9e182e3dd73a/html5/thumbnails/76.jpg)
As n→∞, Students’ t distribution will converge to N(0, 1):
Thm 6. fTn (x)→ fZ (x) =1√2π
e−x22 as n→∞, where Z ∼ N(0, 1).
Proof By Stirling’s formula:
Γ(z) =
√2πz
(ze
)z(1 + O(1/z)) as z →∞
=⇒ limn→∞
Γ( n+1
2
)√
nπ Γ( n
2
) =1√2π
......
36
![Page 77: Math 362: Mathmatical Statistics IImath.emory.edu/~lchen41/teaching/2020_Spring/Chapter-7_full.pdf · Gosset worked at the Guinness Brewery in Dublin, Ireland, and was interested](https://reader036.vdocuments.site/reader036/viewer/2022071005/5fc1e109d75c9e182e3dd73a/html5/thumbnails/77.jpg)
As n→∞, Students’ t distribution will converge to N(0, 1):
Thm 6. fTn (x)→ fZ (x) =1√2π
e−x22 as n→∞, where Z ∼ N(0, 1).
Proof By Stirling’s formula:
Γ(z) =
√2πz
(ze
)z(1 + O(1/z)) as z →∞
=⇒ limn→∞
Γ( n+1
2
)√
nπ Γ( n
2
) =1√2π
......
36
![Page 78: Math 362: Mathmatical Statistics IImath.emory.edu/~lchen41/teaching/2020_Spring/Chapter-7_full.pdf · Gosset worked at the Guinness Brewery in Dublin, Ireland, and was interested](https://reader036.vdocuments.site/reader036/viewer/2022071005/5fc1e109d75c9e182e3dd73a/html5/thumbnails/78.jpg)
Plan
§ 7.1 Introduction
§ 7.2 Comparing Y−µσ/√
n and Y−µS/√
n
§ 7.3 Deriving the Distribution of Y−µS/√
n
§ 7.4 Drawing Inferences About µ
§ 7.5 Drawing Inferences About σ2
37
![Page 79: Math 362: Mathmatical Statistics IImath.emory.edu/~lchen41/teaching/2020_Spring/Chapter-7_full.pdf · Gosset worked at the Guinness Brewery in Dublin, Ireland, and was interested](https://reader036.vdocuments.site/reader036/viewer/2022071005/5fc1e109d75c9e182e3dd73a/html5/thumbnails/79.jpg)
Chapter 7. Inference Based on The Normal Distribution
§ 7.1 Introduction
§ 7.2 Comparing Y−µσ/√
n and Y−µS/√
n
§ 7.3 Deriving the Distribution of Y−µS/√
n
§ 7.4 Drawing Inferences About µ
§ 7.5 Drawing Inferences About σ2
38
![Page 80: Math 362: Mathmatical Statistics IImath.emory.edu/~lchen41/teaching/2020_Spring/Chapter-7_full.pdf · Gosset worked at the Guinness Brewery in Dublin, Ireland, and was interested](https://reader036.vdocuments.site/reader036/viewer/2022071005/5fc1e109d75c9e182e3dd73a/html5/thumbnails/80.jpg)
§ 7.4 Drawing Inferences About µ
Let Y1, · · · ,Yn be a random sample from N(µ, σ2).
Question Find a test statistic Λ in order to test H0 : µ = µ0 v.s. H1 : µ 6= µ0.
Case I. σ2 is known: Λ =Y − µ0
σ/√
n
Case II. σ2 is unknown: Λ =? Λ?=
Y − µ0
s/√
n∼ ?
39
![Page 81: Math 362: Mathmatical Statistics IImath.emory.edu/~lchen41/teaching/2020_Spring/Chapter-7_full.pdf · Gosset worked at the Guinness Brewery in Dublin, Ireland, and was interested](https://reader036.vdocuments.site/reader036/viewer/2022071005/5fc1e109d75c9e182e3dd73a/html5/thumbnails/81.jpg)
§ 7.4 Drawing Inferences About µ
Let Y1, · · · ,Yn be a random sample from N(µ, σ2).
Question Find a test statistic Λ in order to test H0 : µ = µ0 v.s. H1 : µ 6= µ0.
Case I. σ2 is known: Λ =Y − µ0
σ/√
n
Case II. σ2 is unknown: Λ =? Λ?=
Y − µ0
s/√
n∼ ?
39
![Page 82: Math 362: Mathmatical Statistics IImath.emory.edu/~lchen41/teaching/2020_Spring/Chapter-7_full.pdf · Gosset worked at the Guinness Brewery in Dublin, Ireland, and was interested](https://reader036.vdocuments.site/reader036/viewer/2022071005/5fc1e109d75c9e182e3dd73a/html5/thumbnails/82.jpg)
§ 7.4 Drawing Inferences About µ
Let Y1, · · · ,Yn be a random sample from N(µ, σ2).
Question Find a test statistic Λ in order to test H0 : µ = µ0 v.s. H1 : µ 6= µ0.
Case I. σ2 is known: Λ =Y − µ0
σ/√
n
Case II. σ2 is unknown: Λ =? Λ?=
Y − µ0
s/√
n∼ ?
39
![Page 83: Math 362: Mathmatical Statistics IImath.emory.edu/~lchen41/teaching/2020_Spring/Chapter-7_full.pdf · Gosset worked at the Guinness Brewery in Dublin, Ireland, and was interested](https://reader036.vdocuments.site/reader036/viewer/2022071005/5fc1e109d75c9e182e3dd73a/html5/thumbnails/83.jpg)
§ 7.4 Drawing Inferences About µ
Let Y1, · · · ,Yn be a random sample from N(µ, σ2).
Question Find a test statistic Λ in order to test H0 : µ = µ0 v.s. H1 : µ 6= µ0.
Case I. σ2 is known: Λ =Y − µ0
σ/√
n
Case II. σ2 is unknown: Λ =? Λ?=
Y − µ0
s/√
n∼ ?
39
![Page 84: Math 362: Mathmatical Statistics IImath.emory.edu/~lchen41/teaching/2020_Spring/Chapter-7_full.pdf · Gosset worked at the Guinness Brewery in Dublin, Ireland, and was interested](https://reader036.vdocuments.site/reader036/viewer/2022071005/5fc1e109d75c9e182e3dd73a/html5/thumbnails/84.jpg)
§ 7.4 Drawing Inferences About µ
Let Y1, · · · ,Yn be a random sample from N(µ, σ2).
Question Find a test statistic Λ in order to test H0 : µ = µ0 v.s. H1 : µ 6= µ0.
Case I. σ2 is known: Λ =Y − µ0
σ/√
n
Case II. σ2 is unknown: Λ =? Λ?=
Y − µ0
s/√
n∼ ?
39
![Page 85: Math 362: Mathmatical Statistics IImath.emory.edu/~lchen41/teaching/2020_Spring/Chapter-7_full.pdf · Gosset worked at the Guinness Brewery in Dublin, Ireland, and was interested](https://reader036.vdocuments.site/reader036/viewer/2022071005/5fc1e109d75c9e182e3dd73a/html5/thumbnails/85.jpg)
§ 7.4 Drawing Inferences About µ
Let Y1, · · · ,Yn be a random sample from N(µ, σ2).
Question Find a test statistic Λ in order to test H0 : µ = µ0 v.s. H1 : µ 6= µ0.
Case I. σ2 is known: Λ =Y − µ0
σ/√
n
Case II. σ2 is unknown: Λ =? Λ?=
Y − µ0
s/√
n∼ ?
39
![Page 86: Math 362: Mathmatical Statistics IImath.emory.edu/~lchen41/teaching/2020_Spring/Chapter-7_full.pdf · Gosset worked at the Guinness Brewery in Dublin, Ireland, and was interested](https://reader036.vdocuments.site/reader036/viewer/2022071005/5fc1e109d75c9e182e3dd73a/html5/thumbnails/86.jpg)
Summary
A random sample of size n froma normal distribution N(µ, σ2)
σ2 known σ2 unknown
Statistic Z = Y−µσ/√
nTn−1 = Y−µ
S/√
n
Score z = y−µσ/√
nt = y−µ
s/√
n
Table zα tα,n−1
100(1− α)% C.I.(
y − zα/2σ√
n, y + zα/2
σ√n
) (y − tα/2,n−1
s√n, y + tα/2,n−1
s√n
)Test H0 : µ = µ0
H1 : µ > µ0 Reject H0 if z ≥ zα Reject H0 if t ≥ tα,n−1
H1 : µ < µ0 Reject H0 if z ≤ zα Reject H0 if t ≤ tα,n−1
H1 : µ 6= µ0 Reject H0 if |z| ≥ zα/2 Reject H0 if |t | ≥ tα/2,n−1
40
![Page 87: Math 362: Mathmatical Statistics IImath.emory.edu/~lchen41/teaching/2020_Spring/Chapter-7_full.pdf · Gosset worked at the Guinness Brewery in Dublin, Ireland, and was interested](https://reader036.vdocuments.site/reader036/viewer/2022071005/5fc1e109d75c9e182e3dd73a/html5/thumbnails/87.jpg)
Computing s from data
Step 1 a =n∑
i=1
yi
Step 2. b =n∑
i=1
y2i
Step 3. s =
√nb − a2
n(n − 1)
Proof.
s2 =1
n − 1
n∑i=1
(yi − y)2 =n(∑n
i=1 y2i)−(∑n
i=1 yi)2
n(n − 1)
41
![Page 88: Math 362: Mathmatical Statistics IImath.emory.edu/~lchen41/teaching/2020_Spring/Chapter-7_full.pdf · Gosset worked at the Guinness Brewery in Dublin, Ireland, and was interested](https://reader036.vdocuments.site/reader036/viewer/2022071005/5fc1e109d75c9e182e3dd73a/html5/thumbnails/88.jpg)
Computing s from data
Step 1 a =n∑
i=1
yi
Step 2. b =n∑
i=1
y2i
Step 3. s =
√nb − a2
n(n − 1)
Proof.
s2 =1
n − 1
n∑i=1
(yi − y)2 =n(∑n
i=1 y2i)−(∑n
i=1 yi)2
n(n − 1)
41
![Page 89: Math 362: Mathmatical Statistics IImath.emory.edu/~lchen41/teaching/2020_Spring/Chapter-7_full.pdf · Gosset worked at the Guinness Brewery in Dublin, Ireland, and was interested](https://reader036.vdocuments.site/reader036/viewer/2022071005/5fc1e109d75c9e182e3dd73a/html5/thumbnails/89.jpg)
Computing s from data
Step 1 a =n∑
i=1
yi
Step 2. b =n∑
i=1
y2i
Step 3. s =
√nb − a2
n(n − 1)
Proof.
s2 =1
n − 1
n∑i=1
(yi − y)2 =n(∑n
i=1 y2i)−(∑n
i=1 yi)2
n(n − 1)
41
![Page 90: Math 362: Mathmatical Statistics IImath.emory.edu/~lchen41/teaching/2020_Spring/Chapter-7_full.pdf · Gosset worked at the Guinness Brewery in Dublin, Ireland, and was interested](https://reader036.vdocuments.site/reader036/viewer/2022071005/5fc1e109d75c9e182e3dd73a/html5/thumbnails/90.jpg)
Computing s from data
Step 1 a =n∑
i=1
yi
Step 2. b =n∑
i=1
y2i
Step 3. s =
√nb − a2
n(n − 1)
Proof.
s2 =1
n − 1
n∑i=1
(yi − y)2 =n(∑n
i=1 y2i)−(∑n
i=1 yi)2
n(n − 1)
41
![Page 91: Math 362: Mathmatical Statistics IImath.emory.edu/~lchen41/teaching/2020_Spring/Chapter-7_full.pdf · Gosset worked at the Guinness Brewery in Dublin, Ireland, and was interested](https://reader036.vdocuments.site/reader036/viewer/2022071005/5fc1e109d75c9e182e3dd73a/html5/thumbnails/91.jpg)
E.g. 1 How far apart are the bat and the insect when the bat first senses thatinsect is there?
Or, what is the effective range of a bat’s echolocation system?
Answer the question by contruct a 95% C.I.
Sol. ...
42
![Page 92: Math 362: Mathmatical Statistics IImath.emory.edu/~lchen41/teaching/2020_Spring/Chapter-7_full.pdf · Gosset worked at the Guinness Brewery in Dublin, Ireland, and was interested](https://reader036.vdocuments.site/reader036/viewer/2022071005/5fc1e109d75c9e182e3dd73a/html5/thumbnails/92.jpg)
E.g. 1 How far apart are the bat and the insect when the bat first senses thatinsect is there?
Or, what is the effective range of a bat’s echolocation system?
Answer the question by contruct a 95% C.I.
Sol. ...
42
![Page 93: Math 362: Mathmatical Statistics IImath.emory.edu/~lchen41/teaching/2020_Spring/Chapter-7_full.pdf · Gosset worked at the Guinness Brewery in Dublin, Ireland, and was interested](https://reader036.vdocuments.site/reader036/viewer/2022071005/5fc1e109d75c9e182e3dd73a/html5/thumbnails/93.jpg)
E.g. 1 How far apart are the bat and the insect when the bat first senses thatinsect is there?
Or, what is the effective range of a bat’s echolocation system?
Answer the question by contruct a 95% C.I.
Sol. ...
42
![Page 94: Math 362: Mathmatical Statistics IImath.emory.edu/~lchen41/teaching/2020_Spring/Chapter-7_full.pdf · Gosset worked at the Guinness Brewery in Dublin, Ireland, and was interested](https://reader036.vdocuments.site/reader036/viewer/2022071005/5fc1e109d75c9e182e3dd73a/html5/thumbnails/94.jpg)
E.g. 1 How far apart are the bat and the insect when the bat first senses thatinsect is there?
Or, what is the effective range of a bat’s echolocation system?
Answer the question by contruct a 95% C.I.
Sol. ...
42
![Page 95: Math 362: Mathmatical Statistics IImath.emory.edu/~lchen41/teaching/2020_Spring/Chapter-7_full.pdf · Gosset worked at the Guinness Brewery in Dublin, Ireland, and was interested](https://reader036.vdocuments.site/reader036/viewer/2022071005/5fc1e109d75c9e182e3dd73a/html5/thumbnails/95.jpg)
E.g. 2 Bank approval rates for inner-city residents v.s. rural ones.
Approval rate for rurual residents is 62%.
Do bank treat two groups equally? α = 0.05
Sol.H0 : µ = 62 v .s. H1 : µ 6= 62.
43
![Page 96: Math 362: Mathmatical Statistics IImath.emory.edu/~lchen41/teaching/2020_Spring/Chapter-7_full.pdf · Gosset worked at the Guinness Brewery in Dublin, Ireland, and was interested](https://reader036.vdocuments.site/reader036/viewer/2022071005/5fc1e109d75c9e182e3dd73a/html5/thumbnails/96.jpg)
E.g. 2 Bank approval rates for inner-city residents v.s. rural ones.
Approval rate for rurual residents is 62%.
Do bank treat two groups equally? α = 0.05
Sol.H0 : µ = 62 v .s. H1 : µ 6= 62.
43
![Page 97: Math 362: Mathmatical Statistics IImath.emory.edu/~lchen41/teaching/2020_Spring/Chapter-7_full.pdf · Gosset worked at the Guinness Brewery in Dublin, Ireland, and was interested](https://reader036.vdocuments.site/reader036/viewer/2022071005/5fc1e109d75c9e182e3dd73a/html5/thumbnails/97.jpg)
E.g. 2 Bank approval rates for inner-city residents v.s. rural ones.
Approval rate for rurual residents is 62%.
Do bank treat two groups equally? α = 0.05
Sol.H0 : µ = 62 v .s. H1 : µ 6= 62.
43
![Page 98: Math 362: Mathmatical Statistics IImath.emory.edu/~lchen41/teaching/2020_Spring/Chapter-7_full.pdf · Gosset worked at the Guinness Brewery in Dublin, Ireland, and was interested](https://reader036.vdocuments.site/reader036/viewer/2022071005/5fc1e109d75c9e182e3dd73a/html5/thumbnails/98.jpg)
E.g. 2 Bank approval rates for inner-city residents v.s. rural ones.
Approval rate for rurual residents is 62%.
Do bank treat two groups equally? α = 0.05
Sol.H0 : µ = 62 v .s. H1 : µ 6= 62.
43
![Page 99: Math 362: Mathmatical Statistics IImath.emory.edu/~lchen41/teaching/2020_Spring/Chapter-7_full.pdf · Gosset worked at the Guinness Brewery in Dublin, Ireland, and was interested](https://reader036.vdocuments.site/reader036/viewer/2022071005/5fc1e109d75c9e182e3dd73a/html5/thumbnails/99.jpg)
44
![Page 100: Math 362: Mathmatical Statistics IImath.emory.edu/~lchen41/teaching/2020_Spring/Chapter-7_full.pdf · Gosset worked at the Guinness Brewery in Dublin, Ireland, and was interested](https://reader036.vdocuments.site/reader036/viewer/2022071005/5fc1e109d75c9e182e3dd73a/html5/thumbnails/100.jpg)
Plan
§ 7.1 Introduction
§ 7.2 Comparing Y−µσ/√
n and Y−µS/√
n
§ 7.3 Deriving the Distribution of Y−µS/√
n
§ 7.4 Drawing Inferences About µ
§ 7.5 Drawing Inferences About σ2
45
![Page 101: Math 362: Mathmatical Statistics IImath.emory.edu/~lchen41/teaching/2020_Spring/Chapter-7_full.pdf · Gosset worked at the Guinness Brewery in Dublin, Ireland, and was interested](https://reader036.vdocuments.site/reader036/viewer/2022071005/5fc1e109d75c9e182e3dd73a/html5/thumbnails/101.jpg)
Chapter 7. Inference Based on The Normal Distribution
§ 7.1 Introduction
§ 7.2 Comparing Y−µσ/√
n and Y−µS/√
n
§ 7.3 Deriving the Distribution of Y−µS/√
n
§ 7.4 Drawing Inferences About µ
§ 7.5 Drawing Inferences About σ2
46
![Page 102: Math 362: Mathmatical Statistics IImath.emory.edu/~lchen41/teaching/2020_Spring/Chapter-7_full.pdf · Gosset worked at the Guinness Brewery in Dublin, Ireland, and was interested](https://reader036.vdocuments.site/reader036/viewer/2022071005/5fc1e109d75c9e182e3dd73a/html5/thumbnails/102.jpg)
§ 7.5 Drawing Inferences About σ2
For a random sample of size n from N(µ, σ2):
S2 =1
n − 1
n∑i=1
(Yi − Y
)2
⇓
(n − 1)S2
σ2 =1σ2
n∑i=1
(Yi − Y
)2∼ Chi Square(n − 1)
P(χ2α/2,n−1 ≤
(n − 1)S2
σ2 ≤ χ21−α/2,n−1
)= 1− α.
100(1− α)% C.I. for σ2:((n − 1)s2
χ21−α/2,n−1
,(n − 1)s2
χ2α/2,n−1
) 100(1− α)% C.I. for σ:(√(n − 1)s2
χ21−α/2,n−1
,
√(n − 1)s2
χ2α/2,n−1
)
47
![Page 103: Math 362: Mathmatical Statistics IImath.emory.edu/~lchen41/teaching/2020_Spring/Chapter-7_full.pdf · Gosset worked at the Guinness Brewery in Dublin, Ireland, and was interested](https://reader036.vdocuments.site/reader036/viewer/2022071005/5fc1e109d75c9e182e3dd73a/html5/thumbnails/103.jpg)
§ 7.5 Drawing Inferences About σ2
For a random sample of size n from N(µ, σ2):
S2 =1
n − 1
n∑i=1
(Yi − Y
)2
⇓
(n − 1)S2
σ2 =1σ2
n∑i=1
(Yi − Y
)2∼ Chi Square(n − 1)
P(χ2α/2,n−1 ≤
(n − 1)S2
σ2 ≤ χ21−α/2,n−1
)= 1− α.
100(1− α)% C.I. for σ2:((n − 1)s2
χ21−α/2,n−1
,(n − 1)s2
χ2α/2,n−1
) 100(1− α)% C.I. for σ:(√(n − 1)s2
χ21−α/2,n−1
,
√(n − 1)s2
χ2α/2,n−1
)
47
![Page 104: Math 362: Mathmatical Statistics IImath.emory.edu/~lchen41/teaching/2020_Spring/Chapter-7_full.pdf · Gosset worked at the Guinness Brewery in Dublin, Ireland, and was interested](https://reader036.vdocuments.site/reader036/viewer/2022071005/5fc1e109d75c9e182e3dd73a/html5/thumbnails/104.jpg)
§ 7.5 Drawing Inferences About σ2
For a random sample of size n from N(µ, σ2):
S2 =1
n − 1
n∑i=1
(Yi − Y
)2
⇓
(n − 1)S2
σ2 =1σ2
n∑i=1
(Yi − Y
)2∼ Chi Square(n − 1)
P(χ2α/2,n−1 ≤
(n − 1)S2
σ2 ≤ χ21−α/2,n−1
)= 1− α.
100(1− α)% C.I. for σ2:((n − 1)s2
χ21−α/2,n−1
,(n − 1)s2
χ2α/2,n−1
) 100(1− α)% C.I. for σ:(√(n − 1)s2
χ21−α/2,n−1
,
√(n − 1)s2
χ2α/2,n−1
)
47
![Page 105: Math 362: Mathmatical Statistics IImath.emory.edu/~lchen41/teaching/2020_Spring/Chapter-7_full.pdf · Gosset worked at the Guinness Brewery in Dublin, Ireland, and was interested](https://reader036.vdocuments.site/reader036/viewer/2022071005/5fc1e109d75c9e182e3dd73a/html5/thumbnails/105.jpg)
§ 7.5 Drawing Inferences About σ2
For a random sample of size n from N(µ, σ2):
S2 =1
n − 1
n∑i=1
(Yi − Y
)2
⇓
(n − 1)S2
σ2 =1σ2
n∑i=1
(Yi − Y
)2∼ Chi Square(n − 1)
P(χ2α/2,n−1 ≤
(n − 1)S2
σ2 ≤ χ21−α/2,n−1
)= 1− α.
100(1− α)% C.I. for σ2:((n − 1)s2
χ21−α/2,n−1
,(n − 1)s2
χ2α/2,n−1
) 100(1− α)% C.I. for σ:(√(n − 1)s2
χ21−α/2,n−1
,
√(n − 1)s2
χ2α/2,n−1
)
47
![Page 106: Math 362: Mathmatical Statistics IImath.emory.edu/~lchen41/teaching/2020_Spring/Chapter-7_full.pdf · Gosset worked at the Guinness Brewery in Dublin, Ireland, and was interested](https://reader036.vdocuments.site/reader036/viewer/2022071005/5fc1e109d75c9e182e3dd73a/html5/thumbnails/106.jpg)
§ 7.5 Drawing Inferences About σ2
For a random sample of size n from N(µ, σ2):
S2 =1
n − 1
n∑i=1
(Yi − Y
)2
⇓
(n − 1)S2
σ2 =1σ2
n∑i=1
(Yi − Y
)2∼ Chi Square(n − 1)
P(χ2α/2,n−1 ≤
(n − 1)S2
σ2 ≤ χ21−α/2,n−1
)= 1− α.
100(1− α)% C.I. for σ2:((n − 1)s2
χ21−α/2,n−1
,(n − 1)s2
χ2α/2,n−1
) 100(1− α)% C.I. for σ:(√(n − 1)s2
χ21−α/2,n−1
,
√(n − 1)s2
χ2α/2,n−1
)
47
![Page 107: Math 362: Mathmatical Statistics IImath.emory.edu/~lchen41/teaching/2020_Spring/Chapter-7_full.pdf · Gosset worked at the Guinness Brewery in Dublin, Ireland, and was interested](https://reader036.vdocuments.site/reader036/viewer/2022071005/5fc1e109d75c9e182e3dd73a/html5/thumbnails/107.jpg)
§ 7.5 Drawing Inferences About σ2
For a random sample of size n from N(µ, σ2):
S2 =1
n − 1
n∑i=1
(Yi − Y
)2
⇓
(n − 1)S2
σ2 =1σ2
n∑i=1
(Yi − Y
)2∼ Chi Square(n − 1)
P(χ2α/2,n−1 ≤
(n − 1)S2
σ2 ≤ χ21−α/2,n−1
)= 1− α.
100(1− α)% C.I. for σ2:((n − 1)s2
χ21−α/2,n−1
,(n − 1)s2
χ2α/2,n−1
) 100(1− α)% C.I. for σ:(√(n − 1)s2
χ21−α/2,n−1
,
√(n − 1)s2
χ2α/2,n−1
)
47
![Page 108: Math 362: Mathmatical Statistics IImath.emory.edu/~lchen41/teaching/2020_Spring/Chapter-7_full.pdf · Gosset worked at the Guinness Brewery in Dublin, Ireland, and was interested](https://reader036.vdocuments.site/reader036/viewer/2022071005/5fc1e109d75c9e182e3dd73a/html5/thumbnails/108.jpg)
Testing H0 : σ2 = σ20
v.s.
(at the α level of significance)
χ2 = (n−1)s2
σ20
H1 : σ2 < σ20 :
Reject H0 if
χ2 ≤ χ2α,n−1
H1 : σ2 6= σ20 :
Reject H0 if
χ2 ≤ χ2α/2,n−1 or
χ2 ≥ χ21−α/2,n−1
H1 : σ2 > σ20 :
Reject H0 if
χ2 ≥ χ21−α,n−1
48
![Page 109: Math 362: Mathmatical Statistics IImath.emory.edu/~lchen41/teaching/2020_Spring/Chapter-7_full.pdf · Gosset worked at the Guinness Brewery in Dublin, Ireland, and was interested](https://reader036.vdocuments.site/reader036/viewer/2022071005/5fc1e109d75c9e182e3dd73a/html5/thumbnails/109.jpg)
E.g. 1. The width of a confidence interval for σ2 is a function of n and S2:
W =(n − 1)S2
χ2α/2,n−1
− (n − 1)S2
χ21−α/2,n−1
Find the smallest n such that the average width of a 95% C.I. for σ2 is nogreater than 0.8σ2.
Sol. Notice that E[S2] = σ2. Hence, we need to find n s.t.
(n − 1)
(1
χ20.025,n−1
− 1χ2
0.975,n−1
)≤ 0.8.
Trial and error (numerics on R) gives n = 57.
49
![Page 110: Math 362: Mathmatical Statistics IImath.emory.edu/~lchen41/teaching/2020_Spring/Chapter-7_full.pdf · Gosset worked at the Guinness Brewery in Dublin, Ireland, and was interested](https://reader036.vdocuments.site/reader036/viewer/2022071005/5fc1e109d75c9e182e3dd73a/html5/thumbnails/110.jpg)
E.g. 1. The width of a confidence interval for σ2 is a function of n and S2:
W =(n − 1)S2
χ2α/2,n−1
− (n − 1)S2
χ21−α/2,n−1
Find the smallest n such that the average width of a 95% C.I. for σ2 is nogreater than 0.8σ2.
Sol. Notice that E[S2] = σ2. Hence, we need to find n s.t.
(n − 1)
(1
χ20.025,n−1
− 1χ2
0.975,n−1
)≤ 0.8.
Trial and error (numerics on R) gives n = 57.
49
![Page 111: Math 362: Mathmatical Statistics IImath.emory.edu/~lchen41/teaching/2020_Spring/Chapter-7_full.pdf · Gosset worked at the Guinness Brewery in Dublin, Ireland, and was interested](https://reader036.vdocuments.site/reader036/viewer/2022071005/5fc1e109d75c9e182e3dd73a/html5/thumbnails/111.jpg)
E.g. 1. The width of a confidence interval for σ2 is a function of n and S2:
W =(n − 1)S2
χ2α/2,n−1
− (n − 1)S2
χ21−α/2,n−1
Find the smallest n such that the average width of a 95% C.I. for σ2 is nogreater than 0.8σ2.
Sol. Notice that E[S2] = σ2. Hence, we need to find n s.t.
(n − 1)
(1
χ20.025,n−1
− 1χ2
0.975,n−1
)≤ 0.8.
Trial and error (numerics on R) gives n = 57.
49
![Page 112: Math 362: Mathmatical Statistics IImath.emory.edu/~lchen41/teaching/2020_Spring/Chapter-7_full.pdf · Gosset worked at the Guinness Brewery in Dublin, Ireland, and was interested](https://reader036.vdocuments.site/reader036/viewer/2022071005/5fc1e109d75c9e182e3dd73a/html5/thumbnails/112.jpg)
1 > # Example 7.5.12 > n=seq(45,60,1)3 > l=qchisq(0.025,n−1)4 > u=qchisq(0.975,n−1)5 > e=(n−1)∗ (1/l−1/u)6 > m=cbind(n,l,u,e)7 > colnames(m) = c("n",8 + "chi(0.025,n−1)",9 + "chi(0.975,n−1)",
10 + "error ")11 > m12 n chi(0.025,n−1) chi(0.975,n−1) error13 [1,] 45 27.57457 64.20146 0.910330714 [2,] 46 28.36615 65.41016 0.898431215 [3,] 47 29.16005 66.61653 0.886981216 [4,] 48 29.95620 67.82065 0.875953317 [5,] 49 30.75451 69.02259 0.865322418 [6,] 50 31.55492 70.22241 0.855065419 [7,] 51 32.35736 71.42020 0.845161220 [8,] 52 33.16179 72.61599 0.835590121 [9,] 53 33.96813 73.80986 0.826334022 [10,] 54 34.77633 75.00186 0.817376123 [11,] 55 35.58634 76.19205 0.808700824 [12,] 56 36.39811 77.38047 0.800293725 [13,] 57 37.21159 78.56716 0.792141426 [14,] 58 38.02674 79.75219 0.784231327 [15,] 59 38.84351 80.93559 0.776551728 [16,] 60 39.66186 82.11741 0.7690918
50
![Page 113: Math 362: Mathmatical Statistics IImath.emory.edu/~lchen41/teaching/2020_Spring/Chapter-7_full.pdf · Gosset worked at the Guinness Brewery in Dublin, Ireland, and was interested](https://reader036.vdocuments.site/reader036/viewer/2022071005/5fc1e109d75c9e182e3dd73a/html5/thumbnails/113.jpg)
51
![Page 114: Math 362: Mathmatical Statistics IImath.emory.edu/~lchen41/teaching/2020_Spring/Chapter-7_full.pdf · Gosset worked at the Guinness Brewery in Dublin, Ireland, and was interested](https://reader036.vdocuments.site/reader036/viewer/2022071005/5fc1e109d75c9e182e3dd73a/html5/thumbnails/114.jpg)
52