Hypothesis Testing
Hypothesis Testing
Hypothesis Testing
• Case write‐up Grading
Great in summarizing as well as answering the questions.
Basically Acceptable Vs. non‐acceptable
B Late submission penalty (50%)
• Quiz
What is the Statistics? … Population
• In general, “STATISTICS” is the science of gathering, • In statistics, the group we wish to study is called the
describing, and analyzing data population.
Interpreting • A population is:
Analyzing Defined by what the researcher is studying.
Collecting The total number of subjects or things we are interested in
studying.
3 4
Frame … Frame
• In statistics, a frame is a list containing all members of the • Examples of Frames:
population. The frame for the population of the U.S.
• A strict definition of census is a survey that includes all would be a rather long list containing over 300 million names.
elements or units in the frame.
If MIS687 class were the population under consideration,
the class roll would be the frame.
Since a frame is the whole list which contains every members
of the population, it would be much easier to develop when For the population of registered voters,
the population size is small or big? the electoral register would be the frame
5 6
Population Parameter … Population Parameter
• Population parameters are facts about the population. • Examples of Population Parameters:
Parameters are descriptions of the population. 67% of Americans 20 and over are overweight.
A population can have many parameters. 7 out of 10 Americans do not exercise regularly.
Parameters can be in the form of percentages, maximums, 70 million Americans suffer from a sleep disorder.
minimums, or other characteristics.
7 8
Sample … Sample
• A sample is a subset of the population which is used to gain
insight about the population. Samples are used to
represent a larger group, the population.
For example, the percentage of votes a presidential Sample 1
candidate received on Election Day is a parameter.
Sample data is used to try to estimate this population POPULATION
parameter.
Another example is sampling a college campus to represent Sample 2
the population of college students in the U.S.
9 10
What is the Statistics? … What is the Statistics?
• A statistic is a fact or characteristic about a sample.
Question
Population Americans
Sample All college students
Statistics 67% are sleep deprived Estimate Defines
Population
Parameter / Variables
For any given sample, a statistic is a fixed number.
Because there are lots of different samples that can be
drawn from the population, statistics vary depending on the
sample collected. Calculate Collect
Data
Statistics (Sample)
Statistics are used as estimates of population parameters.
11 12
Branches of Statistics … Branches of Statistics …
• Descriptive statistics – the collection, organization, • Example 1:
analysis, and presentation of data. • In a survey of 100 students, 83.2% of students are happy
The emphasis in descriptive statistics is analyzing observed with the food in the cafeteria.
measurements usually from a sample. • Identify the descriptive statistic(s). What inferences can be
To comprehend a large set of data, it must be summarized. made?
Descriptive techniques are the most common statistical
applications. Descriptive statistic?
83.2% of the 100 students surveyed are happy with the food
• Inferential statistics – uses descriptive statistics to in the cafeteria.
estimate population parameters; an educated guess about Possible inference?
the population based on sample data. 83.2% of all students are happy with the food in the cafeteria.
13 14
Branches of Statistics
• Example 2:
• A heart researcher is interested in studying the relationship between diets
which are high in calcium and blood pressure in adult females. The
researcher randomly selects 20 female subjects who have high blood
pressure. Ten subjects are randomly assigned to try a diet which is high in
calcium. The other subjects are assigned to a diet with a standard amount
The Normal
of calcium. After one year the average blood pressures for subjects in
both groups will be measured and compared to decide if diets high in Distribution
calcium decrease the average blood pressure.
• Identify the population. Adult females
Average blood
• What characteristic of the population is being measured? pressure
• Identify the sample. 20 females who have high blood pressure
• Is the purpose of the data collection to perform descriptive or inferential
statistics? Inferential statistics
15 16
Normal Distribution … Normal Distribution …
• Normal Distribution • Symmetric and Bell‐shaped?
A continuous probability distribution for a given random
variable, X, that is completely defined by its mean and variance.
• Properties of a Normal Distribution:
1. A normal curve is symmetric and bell‐shaped.
2. A normal curve is completely defined by its mean, , and
variance, ².
3. The total area under a normal curve equals 1.
4. The x‐axis is a horizontal asymptote for a normal curve.
17 18
Normal Distribution … Normal Distribution …
• Completely Defined by its Mean and Standard Deviation • Total Area Under the Curve = 1
An inflection point is a point on the curve where the curvature of the
line changes. The inflection points are located at ‐ and + .
19 20
Normal Distribution … Normal Distribution …
• The x‐Axis is a Horizontal Asymptote • Example
Determine if the following is a normal distribution
a. Birth weights of 75 babies.
Normal
b. Ages (in year) of 250 students in 10th grade.
No, this would be almost uniform
c. Heights of 100 random adult males.
Normal
d. Frequency of outcomes from rolling a die.
No, because the data is discrete
e. Weights of 50 random fully grown tigers.
Normal
21 22
Normal Distribution … Normal Distribution …
• How Many Normal Curves are there? • Area within One Standard Deviation
The area under the curve and the probability of being within
1
one standard deviation of the mean, µ, equals 0.6826.
23 24
Normal Distribution … Normal Distribution …
• Area within Two Standard Deviations • Area within Three Standard Deviations
The area under the curve and the probability of being within The area under the curve and the probability of being within
2
two standard deviations of the mean, µ, equals 0.9544. three standard deviations ( ± 3σ) of the mean, µ, equals 0.9974.
25 26
Normal Distribution … Normal Distribution …
• Definition • Normal Curves
Normal distribution – a continuous probability density function The mean defines the location and the variance determines the
completely defined by its mean and variance. dispersion.
Below are three different normal curves with different means
1
2 x and identical variances.
2
1
f x e 2
2
27 28
Normal Distribution … Normal Distribution
• Normal Curves • Data from Normal Distributions
Changing the variance parameter can have rather significant As the following three histograms demonstrate, data from a
effects on the shape of the distribution population that is assumed to come from a normal population
Below are two different normal curves with identical means and will more closely represent a bell curve as the sample size n
different variances. grows larger
29 30
Standard Normal Distribution
• A standard normal distribution has the same properties as
the normal distribution; in addition, it has a mean of 0 and
a variance of 1.
The Standard Normal
• Properties of a Standard Normal Distribution
Distribution 1. The standard normal curve is symmetric and bell‐shaped.
2. It is completely defined by its mean and standard deviation,
= 0 and ² = 1.
3. The total area under a standard normal curve equals 1.
4. The x‐axis is a horizontal asymptote for a standard normal
curve.
31 32
Reading a Normal Curve Table … Reading a Normal Curve Table …
• Probability of a Normal Curve • Standard Normal Distribution Table
The probability of a random variable having a value in a given Standard Normal Distribution Table from – to positive z
range is equal to the area under the curve in that region. z 0.00 0.01 0.02 0.03 0.04
0.0 0.5000 0.5040 0.5080 0.5120 0.5160
0.1 0.5398 0.5438 0.5478 0.5517 0.5557
0.2 0.5793 0.5832 0.5871 0.5910 0.5948
0.3 0.6179 0.6217 0.6255 0.6293 0.6331
0.4 0.6554 0.6591 0.6628 0.6664 0.6700
0.5 0.6915 0.6950 0.6985 0.7019 0.7054
0.6 0.7257 0.7291 0.7324 0.7357 0.7389
0.7 0.7580 0.7611 0.7642 0.7673 0.7704
P(X>1) = 0.1587
0.8 0.7881 0.7910 0.7939 0.7967 0.7995
33 34
Reading a Normal Curve Table … Reading a Normal Curve Table …
• Standard Normal Distribution Table • Possible cases of reading the table
1.The standard normal tables reflect a z‐value that is rounded to Area to the Left of z?
two decimal places.
2.The first decimal place of the z‐value is listed down the left‐ Area to the Right of z?
hand column.
4.Where the appropriate row and column intersect, we find the
amount of area under the standard normal curve to the left of Area in the Tails?
that particular z‐value.
35 36
Reading a Normal Curve Table … Reading a Normal Curve Table …
• Area to the Left of z? • Find the area to the left of z
a. z = 1.69
0.9545
b. z = 2.03
1 – 0.9788 = 0.0212
c. z = 0
0.5000
37 38
Reading a Normal Curve Table … Reading a Normal Curve Table …
• Area to the Right of z? • Find the area to the Right of z
a. z = 3.02
1-0.9987 = 0.0013
b. z = 1.70
0.9554
c. z = 0
1-0.5 = 0.5000
39 40
Reading a Normal Curve Table … Reading a Normal Curve Table …
• Area Between z1 and z2? • Find the area Between z1 and z2
a. z1 = 1.16, z2 = 2.31
0.9896-0.8770 = 0.1126
b. z1 = 2.76, z2 = 0.31
0.6217 - (1- 0.9971 = 0.0029) = 0.6188
c. z1 = 3.01, z2 = 1.33
0.0918-0.0013 = 0.0905
41 42
Reading a Normal Curve Table … Example 2
• Area in the Tails? • Calculate the probability that a standard normal random
variable is between 0 and 1.27.
Solution
• Look up the value of 1.27 in the table.
• The table value of .8980 is the area under the curve
left of 1.27.
• So, the answer is .3980 ( = .8980 – 0.5 )
1 ‐ = Answer
43 44
Example 2 Example 3
• Calculate the probability that a standard normal random variable • Calculate the probability that a standard normal random variable
is between −1.08 and 0. is between 1.0 and 2.0.
Solution Solution
• The value −1.08 is not given in the table. • First determine the probability that z is less than 2.0, which the table
• Since the distribution is symmetric, the probability that the random gives as .9772.
variable is between −1.08 and 0 is equal to the probability the random • Then determine the probability that z is less than 1.0, which the table
variable is between 0 and 1.08. gives as .8413.
• The table value of 0.3599 is the area under the curve between 0 and 1.08. • The final step is to subtract the probability z is less than 1.0 from the
probability that z is less than 2.0.
z‐Transformations …
• Definition
z‐Transformation – a transformation of any normal variable into
z‐Transformations a standard normal variable. The z‐transformation is denoted by
z and is given by the formula
x
z .
47 48
z‐Transformations … z‐Transformations …
• Converting to the Standard Normal Curve • Example 1
Given = 40 and = 5, indicate the mean, each of the
inflections points, and where each given value of x will appear
on the curve
x1 = 33 and x2 = 51
Solution:
40
35 45
33 51
49 50
z‐Transformations Example 1 …
• Example 2 • Calculate the probability that a normal random variable
Given = 40 and = 5, calculate the standard score for each x with a mean of 10 and a standard deviation of 20 will lie
value and indicate where each would appear on the standard between 10 and 40.
normal curve
x1 = 33 and x2 = 51
Solution:
0
1 1
1.4 2.2
51 52
Example 1 Example 2 …
Solution • Calculate the probability that a normal random variable
with a mean of 10 and a standard deviation of 20 will be
Applying the z‐transformation yields
greater than 30.
10 10 x 40 10
P 10 X 40 P
20 20
P 0 z 1.5
0.4332
53 54
Example 2 Example 3 …
• Suppose that a national testing service gives a test in which
Solution the results are normally distributed with a mean of 400 and
Applying the z‐transformation yields a standard deviation of 100. If you score a 644 on the test,
what fraction of the students taking the test exceeded your
30 10
P X 30 P z score?
20
Solution
P z 1 Let X = a student’s score on the test.
P z 1 = 1 – P (z<1)
= 1 – .8413
.1587
55 56
Example 3
Solution
The first step is to apply the z‐transformation
P X 644 P z
644 400
100
Developing a
= 1 – P (z<2.44) Hypothesis
= 1 – .8927
.0073
Thus, only 0.73% of the students scored higher than your score of 644.
57
Definitions Hypothesis Test
• In a hypothesis test, the null hypothesis is always given the
• Hypothesis – a theory or premise, usually the claim that benefit of the doubt.
someone is investigating.
• That is, we assume that the null hypothesis is true unless
• Null Hypothesis, H0 – describes the currently accepted there is overwhelming evidence from the sample that goes
value for the population parameter. against the null hypothesis.
• Hypothesis Testing is formalized decision making
• Alternative Hypothesis, Ha – describes the claim that is Choose between REJECTING & FAILING TO REJECT a hypothesis
being tested; the mathematical opposite of the null on the basis of a set of observations
hypothesis.
Rejecting & Accepting – Wrong!!!
• Hypothesis Test – compares the merit of the two • Acceptance implies that the null hypothesis is true.
competing hypotheses by examining the data that is • Failure to reject implies that the data are not sufficiently
collected persuasive for us to prefer the alternative hypothesis over the
null hypothesis
59 60
Developing a Hypothesis ... Developing a Hypothesis …
• To write the null and alternative hypotheses: • Example 1: Determine the null and alternative hypotheses
It is generally accepted among leading educators that
the average student studies no more than 15 hours per
1. Write down the claim
week. In the spring newsletter, a national student
organization claims that its members study more per
2. Write down the mathematical opposite week than the average student
Solution:
3. Assign the null and alternative labels Claim: Members of the national student organization
study more per week than the average student.
μ > 15
Remember that the null hypothesis is the currently accepted Mathematical opposite: μ < 15
value for the population parameter.
H0: μ < 15 Current accepted belief
Ha: μ > 15 Testing hypothesis
61 62
Developing a Hypothesis … Developing a Hypothesis
• Example 2: Determine the null and alternative hypotheses • Example 3: Determine the null and alternative hypotheses
A leading news authority claims that the President’s job Leading authorities have stated that approximately 1% of
approval rating has dropped over the past 3 months. dogs attack people even if unprovoked. The animal rights
Previous polls put the President’s approval rating at a organization believes that 1% is not accurate.
minimum rate of 56%. The President’s chief of staff is
concerned about this claim since it is an election year, and he Solution:
wants to run a test on the claim Claim: 1% is not accurate for the percentage of dogs
Solution: who attack people.
Claim: The President’s approval rating has dropped lower than 56%. p ≠ 0.01
p < 0.56 Mathematical opposite: p = 0.01
Mathematical opposite: p > 0.56
H0: p = 0.01 Current accepted belief
H0: p > 0.56 Current accepted belief
Ha: p ≠ 0.01 Testing hypothesis
Ha: p < 0.56 Testing hypothesis
63 64
Definitions Steps for Hypothesis Testing
• Test Statistic – a component of the criteria used to evaluate 1. State the null and alternative hypotheses
the hypothesis which is calculated from the sample data
gathered 2. Set up the hypothesis test by choosing the test statistic and
determining the values of the test statistic that would lead
• Statistically Significant – Data is said to be statistically to rejecting the null hypothesis
significant if it is unlikely that a sample similar to the one
chosen would occur by chance if the null hypothesis is true
3. Gather data and calculate the necessary sample statistics
• Level of Significance, α – the probability of rejecting a true
null hypothesis. It is also the complement to the level of 4. Draw a conclusion
confidence, α, where α = 1 – c
65 66
Conclusions for a Hypothesis Test … Conclusions for a Hypothesis Test …
1. Reject the null hypothesis. • Example 1 – Determine the conclusion
It is generally accepted among leading educators that the
2. Fail to reject the null hypothesis. average student studies no more than 15 hours per week. In
their spring newsletter, a national student organization claims
that its members study more per week than the national
Once you have decided on a conclusion, a discussion of average. After performing a hypothesis test at the 95% level of
the meaning of this conclusion in terms of the original confidence to evaluate the claim of the organization, the
claim is appropriate. researchers’ conclusion is to reject the null hypothesis. Does
this support the organization's claim?
Claim: Members of the national student organization
study more per week than the average student.
μ > 15
67 68
Conclusions for a Hypothesis Test … Conclusions for a Hypothesis Test …
• Solution 1 • Example 2 – Determine the conclusion
The Board of Education for one large school district uses at least
Claim: Members of the national student organization
study more per week than the average student. 10% as the percentage of high school sophomores considering
μ > 15 dropping out of school. A high school counselor in this district
claims that this percentage is too high. A hypothesis test with
Mathematical opposite: μ < 15 α = 0.02 is performed on the counselor’s claim. The result is to
H0: μ < 15 Current accepted belief fail to reject the null hypothesis. Do the findings support the
Ha: μ > 15 Testing hypothesis counselor’s claim?
After performing a hypothesis test at the 95% level of
confidence to evaluate the claim of the organization, the Claim: The drop out rate of at least 10% is too high.
researchers’ conclusion is to reject the null hypothesis. p < 0.10
Since the null hypothesis was rejected, the organization can
be 95% confident in its claim that its members study more Mathematical opposite: p > 0.10
per week than the national average.
69 70
Conclusions for a Hypothesis Test
• Solution 2
Claim: The drop out rate of at least 10% is too high.
p < 0.10 Reaching a Conclusion
Mathematical opposite: p > 0.10
Since the null hypothesis was not rejected, the evidence is
not strong enough at this level of significance to support the
counselor’s claim.
71
Definitions Reaching a Conclusion …
• Type I Error – rejecting a true null hypothesis; the • Example 1 ‐ Determine if an error was made
probability of making a Type I error is denoted by α. A television executive believes that at least 99% of households own
Inversely related a television set. An intern at her company is given the task of
• Type II Error – failing to reject a false null hypothesis; the testing the claim that the percentage is actually less than 99%. The
probability of making a Type II error is denoted by β. hypothesis test is completed, and based on the sample collected
the intern decides to fail to reject the null hypothesis.
Reality If, in reality, 97.5% of households own a television set, was an error
H0 is true H0 is false made? If so, what type?
73 74
Reaching a Conclusion … Reaching a Conclusion …
• Solution 1
• Example 2 ‐ Determine if an error was made
First state the hypotheses.
Insurance companies commonly use 1000 miles as the average
H0: p > 0.99
number of miles a car is driven per month. One insurance company
Ha: p < 0.99 claims that due to our more mobile society, the average is more
The decision was to fail to reject the null hypothesis. than 1000 miles per month. The insurance company tests their
In reality, the null hypothesis is false since less than 99% of claim with a hypothesis test and decides to reject the null
households own a TV set. hypothesis.
Therefore, the intern failed to reject a false null
Suppose that in reality, the average number of miles a car is driven
hypothesis.
Reality per month is 1500 miles. Was an error made? If so, what type?
This is a Type II error.
H0 is true H0 is false
H0 is rejected Type I error Correct Decision First state the hypotheses.
Decision
H0 is not rejected Correct Decision Type II error H0: μ < 1000
Ha: μ > 1000
75 76
Reaching a Conclusion … Reaching a Conclusion …
• Solution 2 • Example 3 ‐ Determine if an error was made
First state the hypotheses.
A study of the effects of television viewing on children reports that
H0: μ < 1000 children watch an average of 4 hours of television per night. A
Ha: μ > 1000 researcher believes that the average number of hours of television
watched per night by children in her neighborhood is not really 4.
The decision was to reject the null hypothesis. She performs a hypothesis test and rejects the null hypothesis.
In reality, the null hypothesis is false since the average In reality, children in her neighborhood do watch an average of 4
number of miles is more than 1000. hours of television per night. Was an error made? If so, what
Therefore, the decision was to reject a false null type?
hypothesis.
Reality
This is a correct decision. First state the hypotheses.
H0 is true H0 is false
Type I error
H0: μ = 4
H0 is rejected Correct Decision
Decision H : μ ≠ 4
a
H0 is not rejected Correct Decision Type II error
77 78
Reaching a Conclusion …
• Solution 3
First state the hypotheses.
H0: μ = 4
Ha: μ ≠ 4
The decision was to reject the null hypothesis.
In reality, the null hypothesis is true since the children in
her neighborhood watch an average of 4 hours of
television per night.
Therefore, the researcher rejected a true null hypothesis.
This is a Type I error. Reality
H0 is true H0 is false
H0 is rejected Type I error Correct Decision
Decision
H0 is not rejected Correct Decision Type II error
79
Testing a Hypothesis
In the case of n > 30
About a
Population Mean
Definition Steps for Hypothesis Testing …
• Rules for a z‐test for a population mean
o If n > 30, or 1) Define the hypotheses in plain English.
o the standard deviation of the population, σ, is known and the
sample is drawn from a normal population, 2) Select the appropriate statistical measure.
o then by the Central Limit Theorem the test statistic is given by
3) Determine whether the alternative hypothesis should be
x 0
z , where x , one‐sided ( > or < ) or two‐sided ( = or ≠ ).
x n
x is the sample mean, 4) State the hypotheses using the statistical measure that reflects
0 is the hypothesized value of the population mean, the hypotheses under consideration.
and n is the sample size.
5) Specify α, the level of the test.
• If σ is unknown and n > 30, the sample standard deviation, s,
can be used as an approximation of σ. 6) Select the appropriate test statistic
• The test statistic, z, has a standard normal distribution (A test statistic is a quantity calculated from our sample of data)
83 84
Steps for Hypothesis Testing
7) Determine the critical value of the test statistic.
(Critical value is a threshold to which the value of the test statistic in a sample
Matter of > or < ? Ma er of = or ≠ ?
is compared to determine whether or not the null hypothesis is rejected.
Critical value corresponds to a given significance level) One sided Two sided
8) Compute the test statistic using the sample data.
9) Make the decision (Compare Test Statistic and the critical value)
• If Test Statistic > Critical Value Reject H0
• If Test Statistic < Critical Value Fail to reject H0
10) State the conclusion in terms of the original question.
85
Hypothesis Testing … Hypothesis Testing …
• Example 1 • Solution 1
Suppose the national average reading speed for high school 1) H0: Lincoln High School tenth graders are reading at the national
sophomores is 150 words per minute with a standard deviation average.
of 15. A local school board member wants to know if Ha: Lincoln High School tenth graders are not reading at the
sophomore students at Lincoln High School read at a level national average.
different from the national average for tenth graders. The level
of the test is to be set at 0.05. A random sample of 100 tenth 2) μ = average number of words read per minute by Lincoln High
graders has been drawn, and the sample mean is 157 words per School sophomores.
minute.
3) The key word here is “different.” Since the board member is
interested in whether there is any difference, the test is two‐sided.
87 88
Hypothesis Testing … Hypothesis Testing …
• Solution 1 – Contd • Solution 1 – Contd
89 90
Hypothesis Testing … Hypothesis Testing …
• Solution 1 – Contd • Solution 1 – Contd
The rejection regions can be graphed on the real number line,
The z‐test statistic has as follows.
a standard normal distribution.
Then the test statistic can be compared against the rejection
If the null hypothesis is true, regions on the real number line.
95% of the time the value of
the z‐test statistic will be
between ‐1.96 and 1.96.
91 92
Hypothesis Testing … Hypothesis Testing …
• Example 2 • Solution 2
A microprocessor designer has developed a new fabrication 1) H0: The new process does not increase the life of the chip.
process which he believes will increase the usable life of a chip. Ha: The new process does increase the life of the chip.
Currently the usable life is 16,000 hours with a standard
deviation of 2500. Test the hypothesis that the process 2) μ = the mean life of the newly fabricated chips.
increases the usable life of a chip, at the 0.01 level. A sample of
1000 microprocessors is tested and the mean is found to be 3) The key word here is “increase.” Since the goal is to determine if
16,500. Assume the standard deviation of the life of the new there sufficient evidence that the new fabrication process increases
chips will be equal to the standard deviation of the current the usable life of a chip, the test is one‐sided.
chips. 4) H0 : μ < 16,000 Ha : μ > 16,000
5) α = 0.01
93 94
Hypothesis Testing … Hypothesis Testing
• Solution 2 – Contd • Solution 2 – Contd
6) Since n > 30, we can assume that the sampling distribution of x is If the null hypothesis is true, 99% of the time the value of the
approximately normally distributed and therefore we can use the z‐test statistic will be less than 2.33.
z‐test.
7) Since α = 0.01 and the test is one‐sided, za = z0.01 = 2.33.
x 0 16,500 16,000 500
8) z 6.32
2500 2500
n 1000 31.622777
9) Since 6.32 > 2.33 = za, reject the null hypothesis.
10) There is significant evidence at the 0.01 level that the new
fabrication process increases the usable life of a chip.
95 96
Critical Values of the z‐Test Statistic … Critical Value for Z‐Distribution …
• Two sided ( = Two tailed) • If the level of significance is 0.1
Critical Values of the z-Test Statistic for Two-Sided Alternatives Level of Test z
Level of Test Definition of Ordinary Variability z 0.20 1.28
0.20 80% interval around the hypothesized mean 1.28 0.10 1.645
0.10 90% interval around the hypothesized mean 1.645 0.05 1.96 0.90
0.05 95% interval around the hypothesized mean 1.96 0.01 2.575
0.02 98% interval around the hypothesized mean 2.33
0.01 99% interval around the hypothesized mean 2.575 0.05 0.05
z
z = 0
97 98
Critical Value for Z‐Distribution … Critical Values of the z‐Test Statistic …
• If the level of confidence is 0.1,
• One sided ( = One tailed)
Level of Test z Critical Values of the z-Test Statistic for One-Sided Alternatives
0.90
0.20 1.28 Level of Test Definition of Ordinary Variability z
Definition
• Rules for a t‐test for a population mean
o If n < 30, and
o the standard deviation of the population, σ, is unknown and the
sample is drawn from a normal population,
o then the test statistic is given by
P value
x 0 s
t , where sx ,
sx n
x is the sample mean,
0 is the hypothesized value of the population mean,
s is the sample standard deviation,
and n is the sample size.
The test statistic has a t‐distribution with n ‐ 1 degrees of freedom.
103
Definition P‐value …
• P‐value • P‐values allow us to express the strength of our conclusion
The probability of observing a value of the test statistic as to reject or fail to reject the null hypothesis. A smaller P‐
extreme or more extreme than the observed one, assuming the value indicates a rarer test statistic, which implies that the
null hypothesis is true. null hypothesis is less likely to be true.
The p-value is the probability that your null hypothesis
is actually correct • The further a test statistic penetrates a rejection region, the
smaller the P‐value, and the more confidence a researcher
• Decision making with the P‐value can place in the decision to reject the null hypothesis.
If the computed P‐value is less than α, reject the null hypothesis
in favor of the alternative. • On the other hand, if a P‐value is larger than (or equal to) α,
the test statistic is not sufficiently rare to reject the null
If the computed P‐value is greater than or equal to α, fail to
hypothesis. It could have been caused by ordinary
reject the null hypothesis.
sampling variation.
105 106
P‐value … P‐value …
• Less than tests (one‐sided, left‐tailed) – If you are testing • Greater than tests (one‐sided, right‐tailed) – If you are
whether an observed value is significantly less than the testing whether an observed value is significantly greater
hypothesized mean, the P‐value is the probability of than the hypothesized mean, the P‐value is the probability
observing a test statistic less than or equal to the calculated of observing a test statistic greater than or equal to the
value. calculated value.
Example: If the test statistic for a left‐tailed test is z = ‐2.45, Example: If the test statistic for a right‐tailed test is z = 2.12,
then, what is the P‐value? then, what is the P‐value?
Solution Solution
Since it is left‐tailed test, the p‐value for this situation is the Since it is right‐tailed test, the p‐value for this situation is the
probability that z is less than –2.45 = P(z < ‐2.45) = 0.0071. probability that z is greater than 2.12 = P(z > 2.12) = 0.0170
107 108
P‐value … P‐value …
• Not equal tests (two‐sided) – If you are testing whether an • Conclusions for a Hypothesis Testing Using p‐Values
observed value is significantly different from the
hypothesized mean, the P‐value is the probability of 1. If p ≤ α, then reject the null hypothesis.
observing a test statistic whose absolute value is greater
than or equal to the absolute value of the calculated value. 2. If p > α , then fail to reject the null hypothesis.
Thus, to compute the P‐value for a two‐sided test, simply
double the tail probability of the test statistic.
Example: If the test statistic for a two‐tailed test is z = ‐2.56,
then P -value P z 2.56
The P‐value of 0.0104 is the probability
2 P z 2.56
of observing a value of the test statistic
2 0.0052
that is either greater than or equal to 2.56
0.0104.
or less than or equal to ‐2.56 assuming
the null hypothesis is true.
109 110
P‐value … P‐value …
• Example 1 • Example 2
Suppose a hypothesis test about a population mean was conducted The following graph displays a z‐test statistic of 2.03 and the
and the P‐value was calculated to be 0.0156. corresponding P‐value of 0.0212. At a significance level of 0.05,
a) If the level of significance is 0.01, what is the conclusion of the would you reject or fail to reject the null hypothesis?
test?
Solution:
b) If the level of confidence is 90%, what is the conclusion of the
test? Since P‐value = 0.0212 < 0.05 = α
(the significance level), we would
Solutions: reject the null hypothesis.
a. Since P‐value = 0.0156 > 0.01 = α, fail to reject the null hypothesis.
111 112
P‐value Hypothesis Testing …
• Example 3 • Example 1
The following graph illustrates the P‐value for a two‐tailed test Suppose the national average reading speed for high school
where the computed test statistic was z = 2.31. At a significance sophomores is 150 words per minute with a standard deviation
level of 0.05, would you reject or fail to reject the null of 15.
hypothesis? Solution: A local school board member wants to know if sophomore
To calculate the P‐value for a two‐ students at Lincoln High School read at a level different from the
sided test, double the tail national average for tenth graders. The level of the test is to be
probability of the test statistic. set at 0.05.
Thus, P‐value = 0.0104 + 0.0104 A random sample of 100 tenth graders has been drawn, and the
= 0.0208. sample mean is 155 words per minute.
Since P‐value = 0.0208 < 0.05 = α ,
we reject the null hypothesis.
113 114
Hypothesis Testing … Hypothesis Testing …
• Solution 1 • Solution 1 – Contd
1) H0: Lincoln High School tenth graders are reading at the national x 0 155 150 5
7) z 3.33
average. 15 15
Ha: Lincoln High School tenth graders are not reading at the national n 100 10
average. Note that s is used as an approximation of σ.
2) μ = average number of words read per minute by Lincoln High
School sophomores. 8) P -value P z 3.33 2 P z 3.33 2 0.0004 0.0008
3) The key word here is “different.” Since the board member is
interested in whether there is any difference, the test is two‐sided. 9) Since P‐value = 0.0008 < 0.05 = α, reject the null hypothesis.
4) H0 : μ = 150 Ha : μ ≠ 150
10) There is significant evidence at the 0.05 level that tenth graders
5) α = 0.05 at Lincoln High School do not read at the national average.
6) Since n > 30, we can assume that the sampling distribution of x is
approximately normally distributed and therefore we can use the
z‐test.
115 116
Hypothesis Testing … Hypothesis Testing …
• Example 2 • Solution 2
A reporter is investigating a local resident’s claim that gas pumps 1) H0: The average amount of gas dispensed from each pump is at least
typically dispense less gas than the amount purchased. one gallon.
To test the man’s claim, the reporter randomly selects 46 gas Ha: The average amount of gas dispensed from each pump is less than
pumps around town and measures the amount of gas actually one gallon.
dispensed into a can when exactly one gallon is purchased. The 2) μ = the average amount of gas dispensed when one gallon is
purchased.
sample mean amount of gas dispensed is 0.97 gallons with a
3) The key words here are “less than.” Since the reporter is interested in
standard deviation of 0.18 gallons.
whether the pump is dispensing less than one gallon, the test is one‐
Does the data support the man’s claim at the 0.05 significance sided.
level? 4) H0 : μ > 1 Ha : μ < 1
5) α = 0.05
117 118
Hypothesis Testing P‐values for t‐test statistics …
• Solution 2 • Since there is a different t distribution for each number of
6) Since n > 30, we can assume that the sampling distribution of x is degrees of freedom (df), the t table is constructed
approximately normally distributed and therefore we can use the differently from the z tables.
z‐test.
x 0 0.97 1 0.03 • The t table only provides t‐values for frequently used tail
7) z
0.18
0.18
1.13
probabilities, so in most cases the exact value of the t‐test
n 46 6.782330 statistic will NOT be in the table.
Note that s is used as an approximation of s. • When this situation arises, we can find the closest t‐values
8) P -value P z 1.13 0.1292 in the table with the appropriate degrees of freedom which
9) Since P‐value = 0.1292 > 0.05 = α, fail to reject the null surround the test statistic and determine an interval
hypothesis. estimation of the P‐value.
10) There is not significant evidence at the 0.05 level that the average • Exact P‐values for t‐test statistics can be found using a
amount of gas dispensed from each pump is less than one gallon. graphing calculator or statistical software.
119 120
P‐values for t‐test statistics … P‐values for t‐test statistics
• Solution
• Example Degree of freedom? 15
The graph displays a t‐test statistic of 2.236 for a right‐tailed test The value of the test statistic, t = 2.236,
with 15 df. falls between t0.025, 15 = 2.131 and t0.010, 15 = 2.602,
Determine the P‐value.
so P(t > 2.236) is between 0.010 and 0.025.
In interval notation, the P‐value is in the interval (0.010, 0.025).
Using a calculator or statistical software, the exact P‐value can be
found to be 0.0205.
Degrees of Area in One Tail
Freedom t0.100 t0.050 t0.025 t0.010 t0.005
1 3.078 6.314 12.706 31.821 63.657
M M M M M M
121 122
Hypothesis Testing … Hypothesis Testing …
• Solution 1
• Example 1 1) H0: The steel bolts have a mean tensile strength of no more than
The Alexander Bolt Company produces half‐inch A‐class stainless 4000 pounds.
steel bolts. The specified standard is that they have a mean Ha: The steel bolts have a mean tensile strength of more than
tensile strength of more than 4000 pounds. The company is 4000 pounds.
very concerned about quality and wants to be sure that its 2) μ = the mean tensile strength of the steel bolts.
product does not fall below the standard. 3) The key word here is “more.” Since the company is interested if the
A sample of 25 bolts is randomly selected and the mean tensile bolts have a tensile strength of more than a specified value, the test
strength is found to be 4014 pounds, with a standard deviation is one‐sided.
of 20 pounds. 4) H0 : μ < 4000 Ha : μ > 4000
Conduct a hypothesis test to determine if there is overwhelming 5) α = 0.01
evidence at the 0.01 level of significance that the bolts meet the 6) Since n < 30, we will use the t‐test. We assume that the tensile
specified quality standards. strength levels of the bolts are normally distributed.
123 124
Hypothesis Testing … Hypothesis Testing …
• Solution 1 – contd • Example 2
7) t x 0 4014 4000 14 3.5 A sheriff claims that the mean age of prison inmates in his
s 20 20
county is not the same as the mean age of prison inmates in the
n 25 5
state. The mean age of current inmates in the state is 33.7
8) Using the row of the t table for df = 24, we can determine that the years. The sheriff randomly selects the records of 25 inmates in
test statistic, t = 3.5, is greater than t0.005, 24 = 2.797, so P(t > 3.5) is
his county and finds that they have a mean age of 31.3 years
less than 0.005. In interval notation, this means the P‐value is in
with a standard deviation of 8.8 years. Does the data support
the interval (0, 0.005). Using a calculator or statistical software,
the sheriff’s claim at a 0.10 significance level? Assume that the
the exact P‐value can be found to be 0.0009.
age of prison inmates in the county has an approximately
9) Since P‐value = 0.0009 < 0.01 = α, reject the null hypothesis. normal distribution.
10) There is significant evidence at the 0.01 level that the mean Solution:
tensile strength of the steel bolts is greater than 4000 pounds, so 1) H0: The inmates in the county have a mean age of 33.7 years.
the bolts meet the specified quality standards. Ha: The inmates in the county do not have a mean age of 33.7 years.
125 126
Hypothesis Testing … Hypothesis Testing …
• Solution 2 – Contd • Solution 2 – Contd
2) μ = the mean age of all inmates in the county. x 0 31.3 33.7 2.4
7) t s
8.8
8.8
1.364
3) The key words here are “not the same.” Since the sheriff is testing n 25 5
whether the mean age of county inmates is different from the
state average, the test is two‐sided. 8) The test statistic has a t‐distribution
with n ‐ 1 = 24 degrees of freedom.
4) H0 : μ = 33.7 Ha : μ ≠ 33.7 Since the test is two‐tailed, the P‐value
is the probability of getting a t‐test
5) α = 0.10 statistic that is either less than or equal
to ‐1.364 or greater than or equal to
6) Since n < 30, we will use the t‐test. As stated in the 1.364 as shown in the graph.
problem, we assume that the ages of inmates in the county
are normally distributed.
127 128
Hypothesis Testing … Hypothesis Testing …
• Solution 2 – Contd • Solution 2 – Contd
8) Using the row of the t table for df = 24, the absolute value of the test statistic,
|t| = 1.364, falls between t0.100, 24 = 1.318 and t0.050, 24 = 1.711, in other words, 9) The P‐value is in the interval (0.100, 0.200), which means
t 0.100 , 24 = 1.318 < t p‐value/2 , 24 = 1.364 < t 0.050 , 24 = 1.711 P‐value > 0.100 and P‐value < 0.200. Thus, P‐value > α = 0.10, so we
fail to reject the null hypothesis.
ିா
0.05 < < 0.1 10) At the 0.10 level of significance, the data does not support the
ଶ
sheriff’s claim that the mean age of inmates in his county differs
0.05 * 2 < p‐value < 0.1 * 2 0.1 < p‐value < 0.2 from the mean age of inmates in the state.
(so the P‐value, P(|t| > 1.364), is between 2(0.050) = 0.100 and 2(0.100) = 0.200.)
Note that we multiply the area in one tail by 2 since we are conducting a two‐
tailed test.
In interval notation, the P‐value is in the interval (0.100, 0.200). Using a calculator
or statistical software, the exact P‐value can be found to be 0.1852.
129 130
Assignment
• Review & Understand the Lecture Note
• This is critical since Regression (next
week’s content) needs a good
understanding of all the concepts we
learned today.
131