Chapter 9 – Part I

Fundamentals of Hypothesis Testing:

One-Sample Tests
 大綱
1. 假設檢定 (Hypothesis Testing) 概念
2. 操作方法
-critical value approach
-p-value approach
3. 應用
- 估計 population mean
- population std 已知
- population std 未知
- 估計 population proportion
4. 觀念補充
 0. Hypothesis Testing 概念
( 假設檢定 )

目標 : 估計母體某特徵參數 (population parameter)

◦ population mean μ
◦ population proportion π
概念 : 反證法
◦ 先對母體特徵可能值做假設再比對樣本 ( 證據 ) 看有無矛盾
 若矛盾，則 reject 原來假設 (H0)
方法 :
◦ Critical value Approach ( 臨界值法 )
◦ P-value Approach
μ : 台灣人平均年齡 ?
H0: μ=50
H1: μ≠50

H0: μ<=50
H1: μ>50
 What is a Hypothesis?
Ahypothesis is a claim (assertion)
about a
population parameter:
◦ population mean:
Example: The mean monthly cell phone bill
in this city is μ = $42.

◦ population proportion:
Example: The proportion of adults in this
city with cell phones is π = 0.88.
 The Null Hypothesis, H0
States the claim or assertion to be
tested.
H0 : μ
Example: The 30 diameter of a
mean
manufactured bolt is 30mm (
)
Isalways about a population
parameter, not about a sample
H 0 : μ. 30
statistic H 0 : X 30
The Null Hypothesis, H0 (continued)

Begin with the assumption that the

null hypothesis is true.
◦Similar to the notion of innocent
until
proven guilty.

Represents the current belief in a

situation.
Always contains “=“, or “≤”, or “≥”
sign.
May or may not be rejected.
 The Alternative
Hypothesis, H1
Is the opposite of the null hypothesis.
◦ e.g., The mean diameter of a
manufactured bolt is not equal to 30mm
( H1: μ ≠ 30 ).
Challenges the status quo.
Never contains the “=“, or “≤”, or
“≥” sign.
May or may not be proven.
Is generally the hypothesis that the
researcher is trying to prove.
 整理表 : 如何決定 H0 vs. H1

H0( 虛無假設 ) H1( 對立假設 )

(1) = 、 >=, <=( 含等 ≠, <, > ( 沒有等號 )
號)
(2) 想推翻的 想證明我們的主張是對的 ( 樣本證據
( 現狀、歷史紀錄、廠商宣稱 支持的 )
…)
(3) 題目中明確設定的問題 : xx 是
否顯著”異”於、”優”於 、”大”
於 、”低於” yy ?
 The Hypothesis Testing
(continued)

Process
 Suppose the sample mean age was X = 20.

 This
is significantly lower than the claimed
mean population age of 50.
 Ifthe null hypothesis were true, the probability
of getting such a different sample mean would
be very small, so you reject the null hypothesis .

 Inother words, getting a sample mean of 20 is

so unlikely if the population mean was 50  you
conclude that the population mean must not be
50.
 Hypothesis Testing Process for μ

 某研究機構宣稱 : The population mean age is 50.

 想要檢定 (test) 一下是否屬實
 H0: μ = 50,
 H1: μ ≠ 50

Sampling
Distribution of X

X
20 μ = 50 ... then you reject
If it is unlikely that you 假設 H0 真 the null hypothesis
would get a sample ... When in fact this were that μ = 50.
mean of this value ... the population mean…
Test Statistic( 統計量 ) vs.
Critical Values( 臨界值 )
Sampling Distribution of the test statistic
H0 真

Region of Region of
Rejection Rejection
Region of
Non-Rejection

Critical Values

“Too Far Away” From Mean of Sampling Distribution

Risks in Decision Making Using
Hypothesis Testing
 Type I Error:
◦ Reject a true null hypothesis.
◦ A Type I error is a “false alarm.”
◦ The probability of a Type I Error is .
 Called level of significance of the test.
 Set by researcher in advance.
 Type II Error:
◦ Failure to reject a false null hypothesis.
◦ Type II error represents a “missed
opportunity.”
◦ The probability of a Type II Error is β.
決策錯誤
真實狀況

決策 H0 True (H0 真 ) H0 False(H0 假 )

H0 真 Correct Decision Type II Error

(Not Reject Confidence = 1 - α P(Type II Error) = β
H0)
H0 假 Type I Error Correct Decision
(Reject H0) P(Type I Error) = α Power = (1 – β)
 H0: 無罪
決策錯誤 H1: 有罪
真實狀況
決策 H0 True (H0 真 ) H0 False(H0 假 )
實際上無罪 ( 好人 ) 實際上有罪 ( 壞人 )
H0 真 Correct Decision Type II Error
(Not Reject = 實際上有罪 , 判斷無罪
H0)
= 縱容壞人

判斷無罪 ( 好人 )
H0 假 Type I Error Correct Decision
(Reject H0) = 實際上無罪
卻判斷有罪 ( 冤枉好人 )
判斷有罪 ( 壞人 )

檢查了一下 , 目前的 Type I error( 冤枉好人 ) 的確較 Type II error( 縱容

壞人 ) 嚴重 , 因此這樣的 H0,H1 擺放是對的 !
某廠商過去舊藥的治癒率 10% ，該廠商推出了一種新藥並宣稱其治癒率超過 10% ，在收集了新藥的一些實驗數據之後，想做假設
檢定來看看是否新藥的治癒率的確優於舊藥 ? 先隨便立了一下假說
H0: 新的比舊的好 但不確定 H0,H1 有沒有放反 , 所以來看看這樣的假設下的
H1: 新的沒比舊的好 Type I error 有沒有比 Type II error 更嚴重 , 若是
的話 , 則 H0,H1 就放對了 !

真實狀況
決策 H0 True (H0 真 ) H0 False(H0 假 )
實際上新的比舊的好 實際上新的沒比舊的好
H0 真 Correct Decision Type II Error
(Not Reject H0) = 實際上新的沒比舊的好 , 判斷新的比舊的好
判斷新的比舊的好

 讓治癒率比舊藥更差的新藥上市，增加風險 ( 較
嚴重的錯誤 )
H0 假 Type I Error Correct Decision
(Reject H0) = 實際上新的比舊的好 , 卻判斷新的沒比舊的
好
判斷新的沒比舊的好  錯失更好的新藥上市的機會 ( 不會更壞，但也不
會更差 , 算是較輕微的錯誤 !)

由於從目前的 H0,H1 假設會讓 Type II error 比 Type I error 還嚴重 , 所以表示 H0,H1
放反了 ! 因此應把假設更正為 H0: 新的沒比舊的好
H1: 新的比舊的好
 整理表 : 如何決定 H0 vs. H1

H0( 虛無假設 ) H1( 對立假設 )

(1) = 、 >=, <=( 含等 ≠, <, > ( 沒有等號 )
號)
(2) 想推翻的 想證明我們的主張是對的 ( 樣本證據
( 現狀、歷史紀錄、廠商宣稱 支持的 )
…)
(3) xx 是否顯著”異”於、”優”於
、”大”於 、”低於” yy ?

(4) 若要檢查 H0 與 H1 是否有放反，可檢查看看 Type 1

error 是否比 Type 2 error 嚴重，
若是，則 H 與 H 設定正確
Possible Errors in Hypothesis
Test Decision Making
The confidence coefficient (1-α)
is the probability of not rejecting
H0 when it is true.

The confidence level of a

hypothesis test is (1-α)*100%.
The power of a statistical test (1-
β) is the probability of rejecting
H0 when it is false.
 Type I & II Error
Relationship
 Type I and Type II errors cannot happen at
the same time.
 A Type I error can only occur if H0 is true.
 A Type II error can only occur if H0 is false.

If Type I error probability () , then

Type II error probability (β)
 Factors Affecting Type II Error( 補充 )

All else equal,

◦ β when the difference
between hypothesized parameter
and its true value .
◦ β when  .
◦ β when σ .
◦ β when n .
 Level of Significance
and the Rejection Region
H0: μ = 30 Level of significance = a
H1: μ ≠ 30
a /2 a /2

30

Critical values

Rejection Region

This is a two-tail test because there is a rejection region in both tails

具顯著性 (significance) = 有充分證據可以拒絕 H0

Critical Value Approach - Process
- 6 steps

(1) 建立假說 : H0 (null) vs. H1(alternative)

(2) 決定顯著水準 (significant level) 與 n(sample size)

(3) 決定樣本統計量 (sample test statistic): Z or t

(4) 決定 臨界值 (critical values) 與拒絕域 : 查表或電腦
(5) 計算樣本統計量 or ( 用樣本資料 )
(6) 結論 : 樣本統計量落於拒絕域 有充分證據”拒絕 H0”;
若落於非拒絕域  無充分證據拒絕 H0(“ 不拒絕 H0”)
Hypothesis Tests for the Mean

Hypothesis
Tests for 

 Known  Unknown
(Z test) (t test)
 理論基礎 :
Central Limit Theorem – 已知 σ

Population mean μ
Population std σ
As long as the sample size n is large
enough (n>=30)
Distribution of sample mean ~
N(μ, )
Sampling
即 of X
Distribution

X
μ = 50
假設 H0 真
理論基礎 2:
Central Limit Theorem – 未知 σ

Population mean μ
Sample std. S (Population std σ 未知 )
As long as the sample size n is large
enough (n>=30)

Distribution of sample mean ~ tn-

1(μ, )
( or )
理論基礎
Central Limit Theorem – 整理

Population
 mean μ

Population σ 已知 σ 未知 ( 用 S 代替 )
std
As long as the sample size n is large enough
(n>=30)
Distribution of ~ N(μ, )
sample mean

另種寫法
Use Z or t 統計量 ?
Populatio
Population Sample n
STD size distributi
on

n>=3 Z
0
pop. std. Pop. Dis.
σ 已知 Normal Z
n<30 Pop. Dis.
Not ??
Normal
n>=3 t
0
pop. std.
Pop. Dis.
σ 未知 Normal t
n<30 Pop. Dis.
Not ??
Normal
 Z Test of Hypothesis for the
Mean (σ Known)

Hypothesis
Tests for 

σKnown
Known σUnknown
Unknown
(Z test) (t test)
The test statistic is:
X μ
ZSTAT 
σ
n
 Two-Tail Tests
 臨界值 (critical) 有 2 個
 拒絕域 H0: μ = 30
={ 比臨界值更極端的值域 } H1: μ ¹
30
/2 /2

30 X
Reject H0 Do not reject H0 Reject H0

-Zα/2 0 +Zα/2 Z

Lower Upper
critical critical
value value
Critical Value Approach: Example 1

DCOVA
A factory claims that the mean diameter of the
manufactured bolts is 30 mm. A random sample of 100 bolts
are selected to test the claim. The sample mean diameter of
the 100 bolts is 29.84 mm. Assume that population
standard deviation σ = 0.8. Please test whether the
population mean diameter of a manufactured bolt is
significantly different from 30mm at 5% level of
significance.

(1) 建立假說: H0 (null) vs. H1(alternative)

H0: μ = 30
H1: μ ≠ 30  two-tail test
(2) 決定顯著水準 (level of significance)與n(sample size)
Critical Value Approach: Example 1
(3) 決定樣本統計量 (sample test statistic): Z or t
已知 σ=0.8
 n > 30 (CLT)  用 Z test
更精確來說 ~ N(μ, )  ~ N(30, )

H0: μ = 30
H1: μ ¹
30
μ = 30 X
假設 H0 真
Critical Value Approach: Example 1
(4) 決定 臨界值 (critical values) 與拒絕域 : 查表或電腦
a = 0.05  two-tail test  /2 = 0.025
 critical values Z 1- /2 = Z 0.975= 1.96 --> 查表
Z /2 = Z 0.025= -1.96
RR( 拒絕域 )={ 1.96 or -1.96 }  灰色區
or 用 excel 函式 Norm.Inv( 累積機率 , mean,std)=
Norm.Inv(0.975,1,0) 查

/2=0.025
/2=0.025
H0: μ = 30
H1: μ ¹
Reject H0 Do not reject H0 Reject H0 X 30
Z0.025 Z0.975
μ = 30 Z
=-1.96 =1.96
假設 H0 真
 Critical Value Approach: Example 1
(5) 計算樣本統計量 𝑍
( 樣本資料
𝑋  ZSTAT or tSTAT )

X μ 29.84  30  0.16
=29.84 
𝑍
Z STAT 
𝑋 σ

0.8

0.08
 2.0

n 100
~ N(30, )

/2=0.025
/2=0.025
H0: μ = 30
H1: μ ¹
μ = 30 X 30
Z0.025 假設 H0 真 Z0.975
=-2 =-1.96
Z
=1.96
 Critical Value Approach:
Example 1
(6) 結論 : 若 < -1.96 or > 1.96  拒絕 H0; 否則不拒絕 H0.

 /2 = 0.025
 /2 = 0.025

Reject H0 Do
Do not
not reject
reject H
H00 Reject
Reject H
H00

-Zα/2 = -1.96 0 +Zα/2 =+1.96

Here, = -2.0 < -1.96= Zα/2  拒絕 H0

 有充分證據說 mean diameter of a manufactured bolt is
not equal to 30
Critical Value Approach: Example 2
A factory claims that the mean diameter of
the
manufactured bolts is 30 mm. A random
sample of 100 bolts are selected to test the
claim. The sample mean diameter of the 100
bolts is 29.84 mm. Assume that population
standard deviation σ = 0.8. Please test
whether the population mean diameter of a
manufactured bolt is less than 30mm
at 5% level of significance.

(1) 建立假說 : H0 (null) vs. H1(alternative)

H0: μ >= 30
H1: μ < 30
(2) 決定顯著水準 (level of significance) 與 n(sample size)
 = 0.05 , n = 100
Critical Value Approach: Example 3

Q 9.6.4
An article in the San Jose Mercury News stated
that students in the California state university
system take 4.5 years, on average, to finish thei
undergraduate degrees. Suppose you believe
that the mean time is longer.
You conduct a survey of 49 students and obtain
a sample mean of 5.1 years. The population
standard deviation is 1.2 years.
Do the data support your claim at the 1% level?
P-Value Approach-Example 3
p-Value Approach
p-value: Probability of obtaining a
test statistic equal to or more
extreme than the observed sample
value given H0 is true.
◦ p-value : observed level of significance.

◦ the smallest value of  for which H0

can be rejected.
P-value 與  的關係圖
Sampling Distribution of the test statistic

Reject H0 Reject H0
 /2 = 0.025  /2 = 0.025

-Zα/2 +Zα/2
𝑍𝑋 Critical Values

“Too Far Away” From Mean of Sampling Distribution

P-value 與  的關係圖
Sampling Distribution of the test statistic

Reject H0 Reject H0
 /2 = 0.025  /2 = 0.025

p-value

-Zα/2 +Zα/2
𝑍𝑋 Critical Values

“Too Far Away” From Mean of Sampling Distribution

當樣本統計量落於拒絕域內  < 
p-value
p-Value Approach to Testing:
Interpreting the p-value

Compare the p-value with :

◦ If p-value <   reject H0.
◦ If p-value    do not reject H0

Remember

◦ If the p-value is low  reject H0

P-value Approach : Process (5 steps)
DCOVA
(1) 建立假說 : H0 (null) vs. H1(alternative)
(2) 決定顯著水準 (level of significance) 與 n(sample
size)
(3) 決定樣本統計量 (sample test statistic): Z or t
(4) 決定臨界值 (critical values) 與拒絕域 : 查表或電腦
(4) 計算樣本統計量與 p-value( 樣本資料  Z or t P-value)
(5) 結論 : p-value <   reject H0
p-value    not reject H0
P-Value Approach: Example 1
A factory claims that the mean diameter of the
manufactured bolts is 30 mm. A random sample of 100 bolts
are selected to test the claim. The sample mean diameter of
the 100 bolts is 29.84 mm.
Assume that population standard deviation σ = 0.8.
Please test whether the population mean diameter of a
manufactured bolt is significantly different from 30mm
at 5% level of significance.

(1) 建立假說 : H0 (null) vs. H1(alternative)

H0: μ = 30
H1: μ ≠ 30  two-tail test
(2) 決定顯著水準 (level of significance) 與 n(sample
size)
P-Value Approach: Example 1

(3) 決定樣本統計量 (sample test statistic): Z or t

已知 σ=0.8  n > 30 (CLT)  用 Z test
更精確來說 ~ N(μ, )  ~ N(30, )

H0: μ = 30
H1: μ ¹
30
μ = 30 X
假設 H0 真
 P-Value Approach: Example 1

(4) 計算樣本統計量 𝑍與𝑋 P-value(  Z or t P-value)

X μ 29.84  30  0.16
=29.84 
𝑍
Z STAT 
𝑋 σ

0.8

0.08
 2.0

n 100
~ N(30, )
/2=0.025
/2=0.025
下一步求 p-
value??
粉紅面積

=29.84 μ = 30 X
假設 H0 真
=2
Z
=-2
P-Value Approach: Example 1
(4)( 續 ) 計算 P-value : How likely is it to get a ZSTAT of -2
(or something further from the mean (Z=0), in either direction) if
H0 is true?

P-value = = 2 * (1 -
)

0.0456
/2=0.025
/2=0.025

μ = 30 X
假設 H0 真
=2
Z
=-2
P-Value Approach: Example 1

(5) 結論 : 檢查 p-value < α?

◦ p-value = 0.0456 < α = 0.05 
Reject H0.
There is sufficient evidence to
conclude the mean diameter of a
manufactured bolt is not equal to
30mm.
 Two Tail Tests 與 Confidence
Intervals 的關聯
 For X = 29.84, σ = 0.8 and n = 100, the 95%
confidence interval is:
0.8 0.8
29.84 - (1.96) to 29.84  (1.96)
100 100

29.6832 ≤ μ ≤ 29.9968

 Since this interval does not contain the hypothesized

mean (30), we reject the null hypothesis at  = 0.05.
 t Test of Hypothesis for the
Mean (σ Unknown)
 Convert sample statistic ( X ) to a tSTAT test statistic
Hypothesis
Tests for 

σKnown
Known σUnknown
Unknown
(Z test) (t test)
The test statistic is:

X μ
t STAT 
S
n
 Example 2: Two-Tail Test
( Unknown)
The mean cost of a hotel room in New York is said
to be $168 per night. To determine if this is true,
a random sample of 25 hotels is taken and
resulted in an X of $172.50 and an S of $15.40.
(Assume that the population distribution is
Normal.)

Please test whether the mean cost of hotel room

in New York is significant different from $168 at
significant level  of 5%.
Critical Value Approach: Example 2

(1) 建立假說 : H0 (null) vs. H1(alternative)

H0: μ = 168
H1: μ ≠ 168  two-tail test
(2) 決定顯著水準 (level of significance) 與 n(sample
size)
 = 0.05 , n = 25
Critical Value Approach: Example 2

(3) 決定樣本統計量 (sample test statistic): Z or t

未知 σ  以 sample std S=15.4 代替
n < 30  Assume pop. Distr. Normal  ok,
用 t test
更精確來說 ~ tn-1(μ, )  ~ tn-1(168, )
H0: μ =168
H1: μ ¹
168
μ = 168 X
假設 H0 真
Critical Value Approach: Example 2
(4) 決定 臨界值 (critical values) 與拒絕域 : 查表或電腦
a = 0.05  two-tail test   /2 = 0.025
 critical values t df, 1- /2 = t 24, 0.975= 2.0639
t df,  /2 = Z24, 0.025= -2.0639
RR( 拒絕域 )={ 2.0639 or -2.0639 }
 灰色區域

/2=0.025
/2=0.025
H0: μ =168
H1: μ ¹
Reject H0 Do not reject H0 Reject H0 X 168
t24,0.025 t24,0.975
μ = 168 t
=-2.0639 =2.0639
假設 H0 真
查表 (Student’s t Table) or
Excel
Upper Tail Area Excel Function
=t.inv( 累積機率 , 自由度 )
=t.INV(0.975,24)
df .10 .05 .025

24 2.0639  /2 = 0.025

The body of the table

contains t values, not 0 t
probabilities
2.0639
 Critical Value Approach: Example 2
(5) 計算樣本統計量 (𝑡
樣本資料
𝑋  ZSTAT or tSTAT )

X μ 172.50  168
=172.5  𝒕
t STAT 
𝑿 S

15.40
1.46

n 25
~ t24(168, )

/2=0.025
/2=0.025
H0: μ = 168

μ = 168 =172.5 X
H1: μ ¹ 168
t24,0.025 假設 H0 真 t24,0.975
t
=-2.0639 =1.46 =2.0639
 Critical Value Approach:
Example 2
(6) 結論 : 若 < -2.0639 or > 2.0639  拒絕 H0;
否則不拒絕 H0.

 /2 = 0.025
 /2 = 0.025

Reject H0 Do
Do not
not reject
reject H
H00 Reject
Reject H
H00

-tα/2 = -2.0639 0 +tα/2 =+2.0639

Here, = 1.46 < 2.0639= tα/2

 Do not reject H0: insufficient evidence that true mean
cost is different from $168.
P-Value Approach: Example 2
 Two Tail Tests 與 Confidence
Intervals 的關聯
 For X = 172.5, S = 15.40 and n = 25, the 95%
confidence interval for µ is:

172.5 - (2.0639) 15.4/ 25 to 172.5 + (2.0639) 15.4/ 25

166.14 ≤ μ ≤ 178.86

 Since this interval contains the Hypothesized mean (168),

we do not reject the null hypothesis at  = 0.05.
One-Tail Tests ( 單尾檢定 )
下尾檢定 a
H0: μ ≥ 2
μ X
Reject H0 Do not reject H0
H1: μ < 2 -Zα or -tα 0

Critical value

上尾檢定
a
H0: μ ≤ 2
H1: μ > 2
Do not reject H0 Reject H0
Z or t 0 Zα or tα
_
X μ Critical value
 Upper-Tail t Test for Mean
( unknown) - Example
A phone industry manager thinks that
customer monthly cell phone bills have
increased, and now average over $52
per month. The company wishes to test
this claim at 10% significant level. A
random sample of size 25 is selected
with a mean of $53.1 and std. of $10.
(Assume a normal population.)
(1) H0: μ ≤ 52 the mean is not over $52 per month
H1: μ > 52 the mean is greater than $52 per month
(i.e., sufficient evidence exists to support the
manager’s claim)
Critical Value Approach: Example 2
 (2)  = 0.10 and n = 25.
 (3) 決定樣本統計量 (sample test statistic): Z or
t  以 S=10 代替  n < 30  pop. Distr. Normal
未知 σ
 ok, 用 t test
~ t24(μ, )  ~ t24(52, )

H0: μ ≤ 52
H1: μ >
52

μ = 52 X
假設 H0 真
Critical Value Approach: Example 2 ***
(4) 決定 臨界值 (critical values) 與拒絕域 : 查表或電腦
a = 0.1  one-tail test
 critical values t df, 1-  = t 24, 0.9= 1.318
RR( 拒絕域 )={ 1.318}
 單邊灰色區域 查表 or 用 excel 函式計算
=t.inv( 累積機率 , 自由度 )

H0: μ ≤ 52
=0.1
H1: μ >
52
Do not reject H0 Reject H0 X
μ = 52 t24,0.90
t
假設 H0 真 =1.318
Critical Value Approach: Example 2

(5) 計算樣本統計量 𝑡
( 樣本資料
𝑋  ZSTAT or tSTAT )
X  μ 53.1  52
𝑡
t STAT 
𝑋 S

10
0.55
=53.1 
n 25
~ t24(52, )

H0: μ ≤ 52
=0.1
H1: μ >
52
μ = 52 X
假設 H0 真 t24,0.9
t
=0.55 =1.318
 Critical Value Approach:
Example 2
(6) 結論 : 若 > 1.318  拒絕 H0; 否則不拒絕 H0.

 = 0.1

Do
Do not
not reject
reject H
H00 Reject
Reject H
H00

0 +tα/2 =+1.318

Here, = 0.55 < 1.318= tα/2

 Do not reject H0: insufficient evidence that true mean
bill is greater than $52.
P-Value Approach: Example 2
查表 or 用 excel 函式計算
前三步驟相同
=1- t.dis(t 值 , 自由度 , 累積機率 )
(4) 計算 P-value 並與 比較

=0.1
 Do not
reject H0
=0.1

μ = 52 X
假設 H0 真
=0.55 Z