Upload
donald-yum
View
212
Download
0
Embed Size (px)
DESCRIPTION
lecture notes
Citation preview
1
Chapter 9
Hypothesis Testing:
One Sample Tests
David Chow
Oct 2014
2
Learning Objectives
The basic principles of hypothesis testing
Use hypothesis testing to test a mean or
proportion
Underlying assumptions
Potential pitfalls and ethical issues
3
Basic Concepts
4
The Hypothesis
A hypothesis is a claim (assumption) about a
population parameter:
Population Mean
Population Proportion
Eg: The mean weight for
kindergarten kids is μ = 20 kg
Eg: The proportion of retirees with smart phones is π = .68
5
The Process
Claim: The population mean age is of security guards is 50.
Draw a sample and find the sample mean.
Population
Sample
6
The Process
Suppose the sample mean age was X = 20
It is different from the assumption of μ = 50
What can we conclude?
1. Best educated guess, OR
2. Statistical reasoning (hypothesis testing here):
If the hypothesis were true, the probability of getting
such a small X is very small
So the hypothesis is rejected
7
The Process – Graphical Illustration
σx = σ / √ n
μ = 50
If Ho is true … it is unlikely that
you would get a
sample mean of this
value ...
... then you reject
Ho that μ = 50
20
IF this were the
population mean…
X
Sampling Dist of X
8
Terminology: Ho
3μ:H0
The starting point of hypothesis testing is a
hypothesis, usually called the null hypothesis
OR: “the null”, or Ho
Eg: The mean no. of TV sets in US homes is 3
In symbols, it is:
9
Terminology: Ho and H1
How to set up Ho?
1. Ho usually refers to the status quo
2. Ho always has a number, and an equality
Must include any one of “=” , “≤” or “”
Hypothesis testing begins by assuming the null is true
An hypothesis must be stated in pairs, i.e., Ho and H1
H1 is the alternative hypothesis. It is the complement of Ho
Eg: If the null is Ho: μ = 3, the alternative is H1: ____.
10
A Quick Recap
Examples
1. State Ho and H1 on mean body temperature
2. State Ho and H1 for the mean age example
One-tailed (lower-tail)
One-tailed (upper-tail)
Two-tailed
0 0: H
0: aH
0 0: H
0: aH
0 0: H
0: aH
A hypothesis test can be one- or two- tailed The test about must take one of the following three forms (0 is the hypothesized value of ):
11
Test Statistic & Critical Values
Critical Values define “Regions of Rejection”
Distribution of the test statistic (e.g., X)
Region of
Rejection
Region of
Rejection
If the sample mean is far from the assumed population mean, the null is rejected.
How far is “far enough” to reject Ho?
We need critical value(s) for our decision.
12
Level of Significance, α
H0: μ ≥ 50
H1: μ < 50
0
H0: μ ≤ 50
H1: μ > 50
a
a
Represents critical value
Lower-tail test
0
Upper-tail test
Two-tail test
Simple Rule:
Rejection
region ____
0
H0: μ = 50
H1: μ ≠ 50
Claim: The population mean age is 50.
a/2
a/2
13
Errors in Decision Making
Your conclusion to a hypothesis testing is subject to
two potential errors that are different:
Type I error: Wrongly reject a true Ho.
It is equal to a, the probability of drawing an “extreme”
sample when Ho is true.
It is set by researcher in advance.
Type II error: Fails to reject a false Ho.
It is called , which is not selected but computed when n
and a are known.
14
Errors in Decision Making
Possible Hypothesis Test Outcomes
Actual Situation
Decision Ho True Ho False
Do Not
Reject Ho
No Error
Probability 1 - α
Type II Error
Probability β
Reject Ho Type I Error
Probability α
No Error
Probability 1 - β
Eg: A Murder Trial
Ho: Innocence
H1: Guilty
Jury decision
Reject the null (i.e., convicting the defendant), or
Do not reject Ho.
Identify the (potential) errors in this decision.
15
Eg: Medical Test
Medical test
Ho: No cancer
H1: Cancer
For this decision, what exactly are the potential errors?
Type I error
wrong diagnosis of cancer -- unnecessary worry or treatment
Type II error
failure to detect cancer -- patient might miss treatment
_____________
_____________
16
a or : Pick Your Poison?
Given the sample size n and a, can be computed.
Tradeoff: If you want to reduce one type of error, it generally results in increasing the other type of error.
The only way to minimize both types of error is to increase the sample size, but this may be infeasible.
Which error to choose?
Which one leads to a more serious consequences?
(The answer varies on a case-by-case basis.)
Set your a (hence ) accordingly.
The common choice is to set a = 5%.
17
Hypothesis Testing: σ Known
18
Hypothesis Testing: σ Known
Two tail test for the mean (assume σ is known):
Convert sample statistic (X) to test statistic (Z):
Use the Z-table to find the critical Z values, given a specified level of significance α.
Decision Rule: If the test statistic falls in the rejection region, reject Ho ; otherwise do not reject Ho
n
σ
μXZ
19
Hypothesis Testing: σ Known
Do not reject H0 Reject H0 Reject H0
For two-tail tests,
there are two critical
values, and two
regions of rejection. a/2
-Z
0
H0: μ = 3
H1: μ ≠ 3
+Z
a/2
Lower
critical value
Upper
critical value
3
Z
X
20
Eg: Mean Weight
Example: Test the claim that the true mean weight of
chocolate bars manufactured in a factory is 3 ounces.
State the appropriate null and alternative hypotheses:
H0: μ = 3 H1: μ ≠ 3 (two tailed test)
Specify the level of significance
Suppose a = .05 is chosen.
Choose a sample size
Suppose a sample of size n = 100 is selected.
21
Eg: Mean Weight
2.0.08
.16
100
0.8
32.84
n
σ
μXZ
Determine the appropriate technique
σ is known so this is a Z test
σ = 0.8 is known from past company records
Set up the critical values
For a = .05 the critical Z values are ±1.96
Compute the test statistic based on the sample data:
Suppose n = 100 and X = 2.84
So the test statistic is:
22
Eg: Mean Weight
Reject H0 Do not reject H0
Decision: Is the test statistic in the rejection region?
a /2
-Z α/2 = -1.96 0
Reject Ho if Z < -1.96
or Z > 1.96; otherwise
do not reject.
a= 0.05
Upper-tail area = ??
Reject H0
+Zα/2 = +1.96
Here, Z = -2.0 < -1.96, so the test statistic is
in the rejection region
Conclusion in non-
technical term:
Based on the
sample evidence,
the mean weight of
chocolate bars is
not equal to 3.
23
Eg: Volume of Soft-Drink
(2) Test statistic: Z = 10.364
(3) Critical values: Z0.01 = 2.327, (Reject of rejection: Z > 2.327 or Z < -2.327)
(4) Conclusion: Reject Ho as the test statistic falls in the region of rejection.
There is sample evidence to reject the claim of =12.00. We conclude
that 12.00 instead.
ANSWER:
Set a = 2%. The sample has a mean of 12.19oz, and a size of 36. Past
record shows that = 0.11oz. Test the claim that = 12.00oz.
(1) Ho: = 12.00; H1: 12.00.
(2) Test statistic Z = (X – ) / X =
(3) Critical values: Z0.01 =
Hence the region of rejection is:
(4) Finally, the conclusion:
24
Summary: Six Steps
Six Steps of Hypothesis Testing:
1. State the null hypothesis Ho and the alternative H1
2. Choose the level of significance and the sample size n
3. Determine the appropriate statistical technique and the test statistic to use
4. Find the critical values and determine the rejection region(s)
5. Collect data and compute the test statistic from the sample result
6. Compare the test statistic to the critical value:
Reject Ho if the test statistic falls in the rejection region
Otherwise do not reject Ho
Express the decision in non-technical terms
25
Hypothesis Testing: σ Known
p-Value Approach
The p-value is the probability of
obtaining a test statistic equal to or
more extreme ( < or > ) than the
observed sample value given Ho is true
It is also called observed level of
significance
26
Hypothesis Testing: σ Known
p-Value Approach
Convert sample statistic (eg, X) to test statistic (eg, Z
statistic ).
Obtain the p-value from a statistical table.
Compare the p-value with a:
If p-value < a , reject Ho
If p-value a , do not reject Ho If the p-value is
low, Ho must go.
27
Hypothesis Testing: σ Known
p-Value Approach
Example: Mean Weight Again.
Ho: = 3.0; H1: ≠3.0
Sample mean = 2.84, n = 100
.02282.0)P(Z
.02282.0)P(Z
X = 2.84 is translated to a Z
score of Z = -2.0
p-value
=.0228 + .0228 = .0456
.0228
a/2 = .025
-1.96 0
-2.0
Z 1.96
2.0
.0228
a/2 = .025
28
Hypothesis Testing: σ Known
p-Value Approach
Compare the p-value with a
If p-value < a , reject Ho
If p-value a , do not reject Ho
Now p-value = .0456
a is chosen to be 0.05
Since .0456 < .05, reject Ho
.0228
a/2 = .025
-1.96 0
-2.0
Z 1.96
2.0
.0228
a/2 = .025
29
Eg: p-value Approach
1. If you use a 0.05 level of significance in a two-tail hypothesis test, what will you decide if the computed value of the test statistic Z is +2.21?
a. Use the critical value approach.
b. Use the p-value approach.
2. Suppose that in a two-tail hypothesis test, you compute the value of the test statistic Z as -1.38. What is the p-value?
1a. Reject Ho as Z > 1.96
1b. p-value = 2 x (0.01355) = 0.0271
As p-value < 0.05, reject Ho.
2. If Z = -1.38, p-value = 2 x (0.08379) = 0.1676
ANSWER
30
Hypothesis Testing: σ Known
Confidence Interval Connections
100
0.8 (1.96) 2.84 to
100
0.8 (1.96) - 2.84
For X = 2.84, σ = 0.8 and n = 100, the 95%
confidence interval is:
2.6832 ≤ μ ≤ 2.9968
Since this interval does not contain the hypothesized
mean (3.0), you reject the null hypothesis at a = .05
31
Hypothesis Testing: σ Known
One Tail Tests In many cases, the region of rejection is located in one
end of the distribution.
In other words, H1 is focused on one direction only.
There is only one region of rejection, whose area is α.
H0: μ ≥ 3
H1: μ < 3
H0: μ ≤ 3
H1: μ > 3
This is a lower-tail test as H1 is focused
on the lower tail below the mean of 3.
This is an upper-tail test as H1 is focused
on the upper tail above the mean of 3.
32
Eg: Upper Tail Tests
There is only one critical
value, since the rejection
area is in only one tail. Reject Ho
Do not reject Ho
α
μ
Critical value
Similarly, by identifying the correct critical value, you
can construct one-sided confidence intervals.
Eg: For an upper tail test,μ ≤ an upper limit.
33
Eg: Phone Bill
A phone industry manager thinks that customer monthly cell phone bills have increased, now averaging more than $52 per month.
The company wishes to test this claim. Past company records indicate thatσ= $10.
H0: μ ≤ 52 the mean is less than or equal to $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)
Form hypothesis:
34
Eg: Phone Bill
Suppose that a = .10 is chosen for this test.
Find the rejection region:
Reject H0 Do not reject H0
a = .10
Z 0
Reject H0
1-a = .90
35
Eg: Phone Bill
Check the critical value:
Z .07 .09
1.1 .8790 .8810 .8830
1.2 .8980 .9015
1.3 .9147 .9162 .9177 z 0 1.28
.08 a = .10
Critical Value
= 1.28
.90
.8997
.10
.90
36
Eg: Phone Bill
Sample information: n = 64, X = 53.1
=10 was known from past company records
Compute the test statistic:
0.88
64
10
5253.1
n
σ
μXZ
37
Eg: Phone Bill
a = .10
1.28 0
Reject H0
1-a = .90
Z = .88
Now, use the p-value approach to solve the problem.
Decision: Do not reject Ho since Z = 0.88 ≤ 1.28
I.e., There is not sufficient evidence that the mean bill is greater than $52.
38
Hypothesis Testing:
σ Unknown
39
Hypothesis Testing: σ Unknown
If the population standard deviation is unknown, simply replace it by the sample standard deviation S.
Because of this change, you use the t distribution to test Ho.
Check t-table (given αand df = n-1).
All other steps, concepts are the same.
Reminder:
As in the confidence interval chapter, when t-distribution is used, assume the population is approximately normal.
No need to have n > 30 if we assume a normal population.
40
Hypothesis Testing: σ Unknown
Recall that the t test statistic with n-1
degrees of freedom is:
n
S
μXt 1-n
41
Eg: Price Watch
The mean cost of a hotel room in New York City is said to be
$168 per night. A random sample of 25 hotels resulted in
X = $172.50 and S = 15.40. Test at the a = 0.05 level.
(A stem-and-leaf display shows the data are approx. normally distributed )
Ho: μ= 168
H1: μ 168
42
Eg: Price Watch
H0: μ = 168
H1: μ ≠168
α = 0.05
n = 25
is unknown, so use
a t-statistic
Critical Value:
t24 = ± 2.0639
Reject H0 Reject H0
α/2=.025
-t n-1,α/2
Do not reject H0
0
α/2=.025
-2.0639 2.0639
t n-1,α/2
Determine the regions of rejection
43
Eg: Price Watch
a/2=.025
-t n-1,α/2 0
a/2=.025
-2.0639 2.0639
t n-1,α/2
1.46
25
15.40
168172.50
n
S
μXt 1n
Conclusion: Do not reject Ho.
There is not sufficient evidence that true mean cost is different from $168
1.46
44
Hypothesis Testing:
Connection to Confidence Intervals
For X = 172.5, S = 15.40 and n = 25, the 95%
confidence interval is:
166.14 ≤ μ ≤ 178.86
Since this interval contains the hypothesized
mean of 168, you do not reject the null
hypothesis at a = .05
25
15.4 (2.0639) 172.5 to
25
15.4 (2.0639) - 172.5
45
Hypothesis Testing: σ Unknown
It is assumed that the sample statistic comes from
a random sample of a normal distribution.
If the sample size is small (< 30), you should use
a histogram to check the normality assumption.
If the sample size is large, the central limit
theorem applies.
46
Testing Proportion
47
Hypothesis Testing
Proportions
Involves categorical variables
Two possible outcomes
“Success” (possesses a certain characteristic)
“Failure” (does not possesses that characteristic)
Fraction or proportion of the population in
the “success” category is denoted by π
48
Hypothesis Testing
Proportions
sizesample
sampleinsuccessesofnumber
n
Xp
pμn
)(1σ
p
Sample proportion in the success category is denoted by p
When both nπ and n(1-π) are at least 5, p can be
approximated by a normal distribution with mean and
standard deviation
49
Hypothesis Testing
Proportions
The sampling distribution of proportion (p)
is approximately normal, so the test statistic
is a Z value:
n
pZ
)1(
50
Eg: Testing Proportion
A marketing company claims that it receives 8%
responses from its mailing.
To test this claim, a random sample of 500 were
surveyed with 30 responses.
Test at the a = .05 significance level.
First, check:
n π = (500)(.08) = 40
n(1-π) = (500)(.92) = 460
51
Eg: Testing Proportion
H0: π = .08 H1: π ≠ .08
α = .05
n = 500, p = .06
Critical Values: ± 1.96
z 0
Reject Reject
.025 .025
1.96 -1.96
Determine region of rejection
52
Eg: Test for Proportion
Of a sample of 899 home-based businesses, 369 are
owned by females. Want to test if π= 0.50.
Sample proportion p = 369/899, n = 899.
Ho: π= 0.50; H1: π 0.50
Test statistic Z = -5.37
At a = 5%, critical values = 1.96.
Ho is rejected by the sample evidence.
ANSWER
53
Eg: Testing Proportion
Do not reject Ho at a = .05
Test Statistic: Decision:
Conclusion:
There isn’t sufficient
evidence to reject the
company’s claim of 8%
response rate.
1.648
500
.08).08(1
.08.06
n
)(1Z
p
z 0
.025 .025
1.96 -1.96
-1.646
54
Potential Pitfalls and
Ethical Considerations
Use randomly collected data to reduce selection biases
No human subjects without informed consent
Choose the level of significance, α, before data collection
Do not employ “data snooping” to choose between one-tail and two-tail test, or to determine the level of significance
Do not practice “data cleansing” to hide observations that do not support a stated hypothesis
Report all pertinent findings
55
Z or t?
Population Mean (μ)
Z: samp dist normally distributed if σ is known
t: use t-distribution if σ is unknown
Need to assume normal population. But n > 30 is
also acceptable.
Population Proportion (π)
Z: binomial approximated by normal dist
56
More Examples
57
Eg: Mean Waiting Time
Has the mean waiting time in a fast-food restaurant has changed from its previous value of 4.5 minutes?
Past experience shows that the population is normally distributed, with a population standard deviation of 1.2 minutes.
A sample of 25 orders is selected. The sample mean is 5.1 minutes.
The level of significance (a) is 0.05
58
Cont: Critical Value Approach
2.50
25
1.2
5.41.5
n
σ
μXZ
H0: μ = 4.5 H1: μ ≠ 4.5
a = .05
Sample size n = 25
Determine the appropriate technique
The population is normal and σ is known (σ = 1.2) so this is a Z test
Set up the critical values
For a = .05 the critical Z values are ±1.96
Compute the test statistic based on the sample data:
59
Cont: Critical Value Approach
Reject H0 Do not reject H0
Decision: Is the test statistic in the rejection region?
a /2
-Z α/2 = -1.96 0
Decision Rule
Reject Ho if
Z < -1.96 or Z > 1.96;
otherwise do not reject.
a= 0.05
Upper-tail area = 0.025
Reject H0
+Zα/2 = +1.96
Here, Z = 2.50 > 1.96, so the test statistic is
in the rejection region
60
Cont: p-Value Approach
Use the p-value approach to solve the mean waiting time problem.
Again we compute the test statistic of 2.50.
Probability (test statistic ≥ 2.50) = 1-0.9938 = 0.0062
The p-value for this two-tail test =2 x 0.0062 = 0.0124
Decision rule: p < a, reject Ho.
GRAPH:
chosen at 0.05
______
______
61
Eg: Mean Monthly Sales
Want to test if average monthly sales is $120.
Suppose a = 0.05,
X = $112.85, n = 12 and s = $20.80.
Answer
Test statistic t = (X – ) / (s/ n) = -1.19.
Critical values: t0.025, 11 = 2.2010.
The test statistic falls in the region of nonrejection.
Do not reject Ho. We don’t have enough evidence to reject the claim.
Caution: Do not say “accept Ho”.
How about the p-value approach?
ANSWER
Review Questions
Level of significance
True or False? level of significance = α = confidence level
Types of error
In hypothesis testing if the null hypothesis has been rejected when the alternative hypothesis has been true, which error has been committed?
Setting hypotheses
The manager of an automobile dealership is considering a new bonus plan in order to increase sales. Currently, the mean sales rate per salesperson is five automobiles per month. The correct set of hypotheses for testing the effect of the bonus plan is ____
p-value approach
A two-tailed test is performed at 95% confidence. The p-value is 0.09. What is the decision?
Review Questions: Testing μ
A random sample of 16 statistics examinations from a large population was taken.
The average score in the sample was 78.6 with a standard deviation of 8.0.
Want to know if the average grade of the population is significantly more than 75.
Assume the distribution of the population of grades is normal.
Is it a two-tailed test?
Do you use Z- or t- test?
Compute the test statistic
The p-value is between ___
Review Questions: Testing π
A random sample of 100 people was taken.
85 of the people in the sample favored Candidate A.
Want to find out whether or not the proportion of the
population in favor of Candidate A is significantly more
than 80%.
Set up the hypotheses.
What is the test statistic and the p-value?