Anda di halaman 1dari 23

# MAED 602: STATISTICAL METHODS

HYPOTHESIS TESTING

## VON CHRISTOPHER G. CHUA, LPT, MST

Session Objectives
In this fraction of the course on Statistical Methods, graduate students
enrolled in the subject are expected to do the following:

## 1. Discuss in detail hypothesis testing. 4. Describe the steps in hypothesis

2. Define important terms related to testing.
hypothesis testing (e.i. null and 5. Test hypotheses using the traditional
alternative hypotheses, test statistic, method.
critical value and region, type of test,
level of significance)
3. Differentiate Types I and II errors in
hypothesis testing.
This slideshow presentation will be made available through the official course website:
mathbychua.weebly.com.
Fundamentals of
Hypothesis Testing
Understanding the jargons
Hypothesis
In statistics, a hypothesis is a claim or statement It may or
about a property of one or more populations. may not be
true.
Examples of statistical hypotheses:
• The average Filipino adult drinks 800 ml of
carbonated drinks in a week.
• Only ten percent of children who enter Grade 1 will
be able to graduate in college.

## Hypothesis testing aims to make a statistical

conclusion about whether to reject or not reject the
hypothesis.
Hypothesis
The best way to determine if a hypothesis Usually impractical
was true would be to examine the entire
population. So, examine random
samples from the
The mean number of text messages received population
per day by kids aged 15-16 years old in Brgy
A is 24. If sample data is not
consistent with
The mean number of text messages received hypothesis, reject it.
per day by Filipino teenagers is 24.
Statistical
Hypothesis
There are two types of statistical
hypotheses.

## The null hypothesis (𝑯𝟎 ) states that

there is NO statistical significance
between two variables in the
hypothesis.
It is the hypothesis that the researcher
tries to disprove.

## The alternative hypothesis (𝑯𝟏 )

states that there is a statistical
significance between two variables.
Usually what the researcher thinks is
true and is testing.
The hypothesis we want to test is if 𝐻1
is likely true.

## There are two possible outcomes:

• Reject H0 because of sufficient
evidence in the sample in favor or
H1;
• Do not reject H0 because of
insufficient evidence to support H1.

## Note that failure to reject H0 does not

mean the null hypothesis is true. There
is no formal outcome that says “accept
H0." It only means that we do not have
sufficient evidence to support H1.
Stating the null and alternative hypotheses
Coca Cola Bottling Company claims There is NO significant
that the mean volume of coke in cans is difference between the
12 oz. A sample of 200 cans of the soft sample mean and the
drink was randomly selected and company’s claim.
measured in terms of its liquid volume ഥ = 𝟏𝟐 𝒐𝒛
𝑯𝟎 : 𝒙
content. The mean of this sample is
12.38 oz. Is there a significant There is a significant
difference between the sample mean difference between the
obtained and the company’s claim? sample mean and the
company’s claim.
TWO-TAILED HYPOTHESIS TEST ഥ ≠ 𝟏𝟐 𝒐𝒛
𝑯𝟏 : 𝒙
Stating the null and alternative hypotheses
The mean daily income of
The mean daily income of 20 families in the families after the
Brgy A is 250. An NGO recently livelihood program is NOT
conducted a livelihood program for significantly greater than
these families. After three months, the the mean daily income
mean daily income of these families was before it.
ഥ = 𝟐𝟗𝟓
𝑯𝟎 : 𝒙
recorded at 295 pesos. Is the mean
daily income after the livelihood The mean daily income of
program significantly greater than the the families after the
mean daily income of the families livelihood program is
before it? significantly greater than
the mean daily income
before it.
ONE-TAILED HYPOTHESIS TEST ഥ > 𝟐𝟗𝟓
𝑯𝟏 : 𝒙
Two-tailed Hypothesis Test
There is NO significant
difference between the
sample mean and the
ഥ = 𝟏𝟐 𝒐𝒛
𝒙
company’s claim.
ഥ = 𝟏𝟐 𝒐𝒛
𝑯𝟎 : 𝒙
Non rejection region
There is a significant
Critical / rejection difference between the
region sample mean and the
ഥ ≠ 𝟏𝟐 𝒐𝒛
𝒙
company’s claim.
The critical region is the set of all values of the test statistic that ഥ ≠ 𝟏𝟐 𝒐𝒛
𝑯𝟏 : 𝒙
would cause us to reject the null hypothesis.
One-tailed Hypothesis Test
The mean daily income of
the families after the
livelihood program is
NOT significantly greater
than the mean daily
income before it.
ഥ = 𝟐𝟗𝟓
𝑯𝟎 : 𝒙
ഥ = 𝟐𝟗𝟓
𝒙

## Non rejection region

The mean daily income of
the families after the
livelihood program is
significantly greater than
Critical / rejection
region the mean daily income
ഥ > 𝟐𝟗𝟓
𝒙
before it.
ഥ > 𝟐𝟗𝟓
𝑯𝟏 : 𝒙
Test Statistic
The test statistic Claims about the Two Dependent Samples
is a sample population mean, 𝒏 > 𝟑𝟎 ഥ 𝒏
𝒅
statistic or a value ഥ−𝝁
𝒙 𝒕=
𝒛= 𝒏 𝒔
based on the 𝝈
sample data used
Two Independent
in making decision Claims about the Samples
about the population mean, 𝒏 ≤ 𝟑𝟎 ഥ𝟏 − 𝒙
𝒙 ഥ𝟐
rejection of the 𝒕=
ഥ−𝝁
𝒙
𝒏
𝒛=
null hypothesis. 𝒔 𝒔𝟐𝟏 𝒔𝟐𝟐
+
𝒏𝟏 𝒏𝟐
Types of Errors
True State of Nature
The null The null
hypothesis is hypothesis is
true. false.
Type I error
We decide to reject Correct
(rejecting a true
the null hypothesis. Decision
null hypothesis)
Decision Type II error
We decide not to (not rejecting a
Correct
reject the null false null
Decision
hypothesis. hypothesis)
Two-tailed Hypothesis Test

## The probability of rejecting In two-tailed tests, the level

the null hypothesis when it of significance level, 𝜶 is
is true (Type I error) is divided equally between
called the significance the two tails that constitute
level, denoted by the the critical region.
symbol 𝜶. This value is
typically predetermined.

## 𝟎. 𝟎𝟐𝟓 Critical / rejection 𝟎. 𝟎𝟐𝟓

region
𝜶 = 𝟎. 𝟎𝟓
One-tailed Hypothesis Test

## One-tailed hypothesis tests In one-tailed tests, the level

can be left tailed or right of significance level, 𝜶 is
tailed depending on which taken only on the tail that
tail the critical region is contains the critical region.
located.

## Non rejection region

Critical / rejection 𝟎. 𝟎𝟓
region
𝜶 = 𝟎. 𝟎𝟓
Conclusions in Hypothesis Testing
Start

Does the
original claim “There is sufficient evidence to warrant
No Reject Yes
contain the the rejection of the claim that … (original
Original 𝐻0 ? (Reject 𝐻0 )
condition of claim claim)”
equality? contains
equality and “There is no sufficient evidence to
becomes 𝐻0 No
(Do not reject 𝐻0 )
warrant the rejection of the claim that …
(original claim)”

No Reject Yes
“The sample data support that …
Original claim does not 𝐻0 ? (Reject 𝐻0 )
contain equality and (original claim)”
becomes 𝐻1

## No “There is no sufficient sample evidence

(Do not reject 𝐻0 ) to support the claim that … (original
claim)”
Steps in Hypothesis
Testing
Going through the process
1. State the null and alternative hypotheses.
Express the hypotheses in symbols.
2. Determine the appropriate test statistic and the
sampling distribution.
3. Identify or select the significance level and
degree of freedom.
4. Determine the critical values and the critical
region and formulate the decision rule.
Traditional 5. Compute for the test statistic.
Method of
6. Decide whether to reject or not to reject the null
Hypothesis hypothesis. Draw a conclusion.
Testing
t-test for One Sample
There is no significant difference From the t-table, we identify
A local gym advertises that with between the mean weight loss of the region to be greater than
their workout plan “you will lose the sample and the local gym’s 2.064.
5 kg of body mass in a month”. A claim of a 5-kg weight loss. Decision rule: Reject 𝐻0 if
𝐻0 : 𝑥ҧ = 5 𝑘𝑔 𝑡𝑐 > 2.064
consumers group wants to test There is a significant difference 4.21 − 5
𝒕𝒄 = 25 = −𝟓. 𝟎𝟔
this claim by getting data on 25 between the mean weight loss of 0.78
the sample and the local gym’s Since 𝑡𝑐 > 2.064, we reject the
people who went through the
claim of a 5-kg weight loss. null hypothesis.
program and found out that the 𝐻1 : 𝑥ҧ ≠ 5 𝑘𝑔
average weight loss of this There is sufficient evidence to
Since 𝑛 < 30, test statistic will be warrant the rejection of the
sample is 4.21 kg with a
obtained through two-tailed t- claim that the mean weight loss
standard deviation of 0.78 kg. test for one sample. of the sample is not
Test the claim of the gym at 𝛼 = 𝒕=
ഥ−𝝁
𝒙
𝒏
significantly different the local
𝝈 gym’s claim of a 5-kg weight
5%.
loss.
Set 𝛼 = 0.05, 𝑑𝑓 = 24
t-test for One Sample
The mean rating of the From the t-table, we identify
As a means to assess its faculty, probationary instructors is not the region to be greater than
the Graduate School of ESSU significantly greater than the 1.706.
conducts a faculty evaluation mean rating of all faculty Decision rule: Reject 𝐻0 if 𝑡𝑐 >
members. 𝐻0 : 𝑥ҧ = 3.39 1.706
before the end of every term. The mean rating of the
For the current semester, the probationary instructors is
𝒕𝒄 =
3.76 − 3.39
27 = 𝟐. 𝟐𝟏𝟎
average rating of all instructors significantly greater than the 0.87
mean rating of all faculty
is 3.39 with a standard deviation members. 𝐻1 : 𝑥ҧ > 3.39 Since 𝑡𝑐 > 1.706, we reject the
of 0.87. On the other hand, null hypothesis.
twenty-seven newly hired Since 𝑛 < 30, test statistic will be
obtained through one-tailed t- There is sufficient evidence to
probationary instructors got a test for one sample. warrant the rejection of the
mean rating of 3.76. Based from 𝒕=
ഥ−𝝁
𝒙
𝒏
claim that mean rating of the
probationary instructors is not
the data, did the probationary 𝝈
significantly greater than the
instructors get a better rating Set 𝛼 = 0.05, 𝑑𝑓 = 26 mean rating of all faculty
than the faculty as a whole? members.
z-test for One Sample
There is no significant difference From the z-table, we identify
A company which manufactures between the mean of the the region that is 0.025 from
battery-operated toy cars randomly selected sample and each tail or 0.475 from the
claims that its products have a the company’s claim. 𝐻0 : 𝑥ҧ = 5 mean, 𝑧 = 1.96.
There is a significant difference Decision rule: Reject 𝐻0 if
mean life span of 5 years with a between the mean of the 𝑧𝑐 > 1.96
standard deviation of 2 years. A randomly selected sample and
sample of 40 toys was tested the company’s claim. 𝐻1 : 𝑥ҧ ≠ 5 3−5
𝒛𝒄 = 40 = −𝟔. 𝟑𝟐
2
and found to have a mean life Since 𝑛 > 30, test statistic will be
span of only 3 years. Using a 5 obtained through two-tailed z- Since 𝑧𝑐 > 1.96, we reject the
test for one sample. null hypothesis.
percent level of significance, ഥ−𝝁
𝒙
determine if there is a 𝒛=
𝝈
𝒏
There is sufficient evidence to
significant difference between warrant the rejection of the
𝛼 = 0.05 claim that the mean life span of
the mean of the randomly the company’s battery-
selected sample and the operated toy cars is 5 years.
company’s claim.
Test concerning two dependent samples
A professor who teaches Algebra offered his students a two-hour lecture on Math anxiety and
ways to overcome it. The following table shows the test score in Algebra of seven students before
and after they attended the lecture. Test at 2.5% level of significance if attending the lecture
helped improve the score in Algebra.

## Student Allan Bobby Carlo Dante Efren Frank Gabby

Before 56 69 48 74 65 71 58
After 62 73 44 85 71 70 69
d
Test concerning two independent samples
Filipino freshmen high school From the z-table, we identify the
A local researcher studied the
students did not perform better region that is 0.05 from each tail or
Mathematics achievement of in Mathematics than their 0.45 from the mean, 𝑧 = 1.645
Amerasians and Filipinos. According to Amerasian counterparts. Decision rule: Reject 𝐻0 if
the study, the mean score on a 𝐻0 : 𝑥𝐹 = 𝑥𝐴 𝑧𝑐 > 1.645
mathematics test given to 120 Filipino freshmen high school
Amerasian freshmen high school students performed better in 𝟔𝟖. 𝟗 − 𝟔𝟖
𝒛𝒄 = = 𝟏. 𝟏𝟔𝟐
students was 68 with a standard Mathematics than their 𝟕. 𝟖 𝟐 𝟓. 𝟔 𝟐
deviation of 5.6. The same test was Amerasian counterparts. +
𝟏𝟖𝟎 𝟏𝟐𝟎
given to 180 Filipino freshmen high 𝐻1 : 𝑥𝐹 > 𝑥𝐴
school students and the mean score Since𝑧𝑐 < 1.645, we do not reject the
Test statistic will be obtained null hypothesis.
obtained was 68.9 with a standard through one-tailed z-test for two
deviation of 7.8. Using the 5% level of independent samples. There is no sufficient evidence to
significance, does the data show that ഥ𝟏 − 𝒙
𝒙 ഥ𝟐 warrant the rejection of the claim
Filipino freshmen high school students 𝒛 =
𝟐 𝟐 that the Filipino freshmen high
performed better in Mathematics than 𝒔 𝒔
𝟏
+ 𝟐 school students did not perform
𝒏𝟏 𝒏𝟐
their Amerasian counterparts? better in Mathematics than their
Amerasian counterparts.
𝛼 = 0.05