MODULE 2

INFERENTIAL STATISTICS – HYPOTHESIS TESTING

1. to differentiate null from alternative hypothesis

2. to formulate the null and alternative hypotheses for a given
research problem

3. to compare and contrast directional and non-directional test

4. to illustrate the steps in hypothesis testing

Basic statistics courses include lessons on summarizing and organizing and

describing a set of data, or a frequency distribution in terms of how do the scores
typically look like (measures of central tendency such as the mean, median and
mode) or how much they spread out from high to low scores (measures of
variability like range, variance and standard deviation). However, analysis goes
beyond this purpose. Many problems require that we decide whether or not a
statement about the mean or the standard deviation of a sample set of scores is
true or false. In statistics, this statement is called hypothesis and the decision
making about whether this statement is true or false is called hypothesis testing.

The Hypothesis

A hypothesis is basically a statement about the target population.

Hypotheses are formulated as a result of years of experience, observations and
researches. However, due to many factors, the whole population cannot be
studied. Instead, we take or select a random sample from the population being
studied and from the data gathered, we decide whether the hypothesis is true or
false.
A statistical hypothesis is an assertion or conjecture concerning one or
more population. It is a preconceived idea about the values of the population
parameters which can be verified through statistical tests. This conjecture may or
may not be true. There are two forms of statistical hypothesis: the null
hypothesis and the alternative hypothesis.

The Null hypothesis (denoted by Ho) is a statistical hypothesis that states

that there is no difference between a parameter and a specific value or there is no
difference between two parameters. This is the hypothesis to be tested and it
represents what the researcher doubts to be true.
The Alternative hypothesis (denoted by Ha) is a statistical hypothesis that
states a specific difference between a parameter and a specific value or that there
is a difference between two parameters. This is the operational statement of the
theory that the researcher believes to be true and aims to prove.

Example1. Suppose we are interested in the average performance of a group of

students in an English test. Specifically, we are interested in deciding (hypothesis
testing) whether or not the mean performance (denoted by the Greek small letter
Mu, μ) of the students is say for example, equal to 85 .

The Null hypothesis (Ho) is : μ = 85 . We are stating that there is no difference

between the mean score of the students and 85

The Alternative hypothesis (Ha) is μ ≠ 85. We are stating that there is a difference
between the mean scores of the students and 85.

The decision is either to retain or reject the null hypothesis. If the

information or data that we gathered about the performance of students in the test
(mean score in the English test) is consistent with the null hypothesis, then we
conclude that the null hypothesis is true which will lead us to retain the null
hypothesis. On the other hand, if the information or data that we gathered is
inconsistent with the null hypothesis, then we would conclude that the null
hypothesis is false which will lead us to reject the null hypothesis.
 As shown in the two hypotheses, the alternative hypothesis is a statement
that contradicts or disagrees with the null hypothesis. If the null hypothesis is
rejected (meaning that the statement indicating there is no difference between the
parameter and the specified value is FALSE), then the alternative hypothesis is
the conclusion or accepted (meaning that the statement indicating that there is
difference between the parameter and the specified value is TRUE).

Example 2 . Suppose we are interested to examine the pretest and posttest Math
performance of a sample of students. Specifically, we are interested in deciding
whether or not the mean performance of the students in the pretest and posttest
are the same..

The Null hypothesis (Ho) is : μ pretest= μ posttest . We are stating that there is no
difference between the mean pretest score and mean posttest score.

The Alternative hypothesis (Ha) is: μ pretest ≠ μ posttest . We are stating that there is
a difference between the mean pretest score and mean posttest score.

Example 3. Suppose we are interested to compare the mean scores of two groups
of students from two different Math classes.

The Null hypothesis (Ho) is : μ Group 1 = μ Group 2 . We are stating that there is no
difference between the mean scores of Group 1 and Group 2.

The Alternative hypothesis (Ha) is: μ Group 1 ≠ μ Group 2 . We are stating that there
is a difference between the mean scores of Group 1 and Group 2.
Directional and Non-directional Tests of Hypothesis

A test of hypothesis is of two types, a one-tailed (directional) test of

hypothesis and a two-tailed (non-directional) test of hypothesis.

A one-tailed test of hypothesis occurs when the research has an a priori

expectation about the sample value he expects to observe. The researcher wants
to test whether a mean is larger or smaller than the hypothesized population mean.

A two-tailed test of hypothesis occurs when the alternative hypothesis

does not specify a directional difference for the parameter of interest. This is
appropriate when the researcher does not have a priori expectation regarding the
value he expects to observe in the sample. The researcher wants to see if the
sample mean is different from the hypothesized population mean, in either
direction.
Example: A classroom teacher hypothesizes that using lecture method in teaching
a specific lesson is the same as teaching it using demonstration method in terms
of students’ learning. This means that students who were taught using the lecture
method would perform similarly (say on a test) than those who were taught using
demonstration method

The null hypothesis is (Ho) μ lecture = μ demonstration . This means that the performance
of the students who were taught using both methods of teaching are the same, or
not different.

There are three possible alternative hypotheses (Ha) to this null hypothesis:
1. μ lecture ≠ μ demonstration The mean score of the students using lecture method
is not equal to the mean score using demonstration
method
2. μ lecture > μ demonstration The mean score of the students using lecture method
is higher than the mean score using demonstration
method
3. μ lecture < μ demonstration The mean score of the students using lecture method
is lower than the mean score using demonstration
method

The first alternative hypothesis allows for the possibility that either
meanlecture is greater than or less than meandemonstration. This is a non-directional
(two-tailed) test. Whereas the second alternative hypothesis is specific that the
meanlecture is greater than meandemonstration – an example of a directional (one-tailed)
test. The third alternative hypothesis is also a directional (one-tailed) test.

The null and alternative hypotheses are stated together as shown:

Two-tailed One-tailed (right) One-tailed (left)
Null Hypothesis (Ho) = ≤ ≥
Alternate Hypothesis (Ha) ≠ > <
Example 1. A teacher thinks that using a module will enhance the performance of
the students in Statistics. In the past, the average grades of students was 75.
Ho : μ ≤ 75. The null hypothesis is one-tailed (right) .The null hypothesis
is that the performance will not be enhanced, so it could the same or
even lower (≤)
Ha: μ > 75. The alternative hypothesis is one-tailed (right) because he
expects that the performance will increase. The alternate hypothesis
is that the performance will be enhanced (>).

Example 2. A psychologist thinks that a therapy will reduce the anxiety of patients
suffering from insomnia. In the past, the average levels of anxiety was 4.5 ( on a
scale of 1-5)
Ho : μ ≥ 75. The null hypothesis is one-tailed (left) .The null hypothesis
is that the anxiety will not be reduced, so it could the same or even
higher (≥)
Ha: μ < 75. The alternative hypothesis is one-tailed (left) because he
expects that the anxiety will decrease. The alternate hypothesis is
that the anxiety will be reduced (<).

Example 3. A research thinks that advertising in radio or print will change the
buying preference of teenagers for a certain perfume. However, the researcher is
not sure whether the sales will increase or decrease after the advertisement. In
the past, the average sales was P50, 000.

Ho: μ = 50,000 ( The average sales after the advertisement does not differ
from 50,000).
Ha : μ ≠ 50,000 ( The average sales after the advertisement is different
from 50,000). It could be higher or lower, thus a directional (two-
tailed) test.
Types of Error

Whichever hypothesis (the null or alternative) is correct, there is always a

chance that the sampling error will cause the researcher to reject that hypothesis.
We may reject a null hypothesis that is true or we may accept a null hypothesis
that is false. These two types of mistakes in statistical hypothesis testing are called
Type I error and Type II error.

A Type I error is incurred if a researcher rejects a null hypothesis when in

fact it is true. The probability of committing a Type I error is denoted by the Greek
letter alpha (α).
A Type II error is incurred when the researcher retains the null hypothesis
when in fact it is false. The probability of committing a Type II error is denoted by
the Greek letter beta (β).

A correct decision is made when the null hypothesis is retained when it is

really true, and therefore should be retained. The probability of making this correct
decision is 1 – α. That is why the adopted level of significance in conducting a
test of hypothesis is 0.05 and the probability of correctly accepting a true
hypothesis is 0.95.
A second correct decision is made when the null hypothesis is rejected
when it is really false, and therefore should be rejected. The probability of making
this correct decision is 1 – β.

The level of significance is the maximum probability of committing a Type

I error. Researchers generally agree on using three arbitrary significance: 0.10,
0.05 and 0.01 level. A level of significance of 0.05 or 5% means that there is 5%
chance of rejecting a true null hypothesis. It also means that when we repeat
sampling from a given population of interest, the probability of getting a result
similar to the one presently observed is 95%, and the probability of getting a
different result is only 5%. The same is true for a significance level of 0.01. The
probability of getting the same result is 99% while the probability of getting a
different result is only 1%. Significance level indicates that the probability of getting
the result due to plain chance is equal to or less than the alpha level (α). The
choice of alpha level depends on the consequences associated with making a
Type I error. The smaller the value of the significance level, the smaller will be the
rejection region.
It should be noted that the α and β are inversely related for a fixed sample
size (n). To reduce both error, there is a need to increase the sample size.

The table summarizes the two types of error.

Null Hypothesis
Decision True False
Reject Ho Type I Correct decision
Accept Ho Correct decision Type II

Using court law decisions to illustrate it further, Type I error is incurred when
an innocent person is convicted guilty of a crime he actually did not commit. On
the other hand, Type II error is incurred when a guilty person is declared innocent
for a crime he actually committed. A correct decision is made when a person who
actually did not commit the crime is declared innocent, or when a person who
actually committed the crime is declared guilty.

Another illustration is on the smoke detectors set up on buildings. If there

is no fire, the smoke detector is either in alarm state or no alarm state.
Alternatively, if there is a fire, the smoke detector is either in alarm state or no
alarm state (i.e. faulty batteries). If the smoke detector alarms even when there is
no fire, then Type I error is incurred. If the smoke detector does not alarm when
there is fire, Type II error is incurred. The smoke detector is efficient (it works well)
if it alarms when there is fire, or if it does alarm when there is no fire.
Steps in Hypothesis Testing

Step 1. Formulate the null and alternative hypotheses.

The null hypothesis is usually that the observations are the result purely of
chance or the hypothesis of no effect.

The alternative hypothesis is that there is a real effect, or that the

observations are the result of this real effect plus chance variations.

Example: A researcher wants to determine whether boys experience the same

levels of test anxiety than girls.

Step 1.

Null hypothesis: There is no difference on the test anxiety between boys

and girls.
Ho : μ anxiety boys = μ anxiety girls)

Alternative hypotheses:

a. There is difference on the test anxiety between boys and girls (two-tailed
test).
Ha : μ anxiety boys ≠ μ anxiety girls

b. The test anxiety of boys is higher than girls (one-tailed test, right tail).
Ha : μ anxiety boys > μ anxiety girls

c. The test anxiety of boys is lower than girls. (one- tailed test, left tail).
Ha : μ anxiety boys < μ anxiety girls

Step 2. Set the level of significance

The level of significance refers to the probability level that is considered

such that the hypothesis being tested warrants support. In social and educational
researches, the level of significance is usually set at 0.05. The level of significance
gives the probability of incorrectly rejecting a true null hypothesis (Type I error).
Hence, this probability should be small.
Step 3. Select the appropriate test statistics

Given a statistical hypothesis, there is a specific and appropriate test

statistics for it. The choice of the statistical tests depend on the following:

a. level of measurement – whether the level of measurement is nominal,

ordinal, interval or ratio

b. objective of the study – whether the researcher would like to compare

means, or to compare two groups , or more than two groups

c. design of the study – whether case control, cohort , cross sectional, and
so on.

d. whether the samples are related (dependent) or independent – In

independent sample, the probability of selection of sample/s in one
group in not affected by the selection in the other group/s.

e. assumptions on the distribution of the data

1. Parametric tests – can be used when the assumptions about the

parameters in the population like normality, independence and
homogeneity hold true.

2. Nonparametric tests –can be used when the assumptions for the

use of parametric tests are questionable in the data. These tests
have fewer and less stringent assumptions about the population
parameters.

In our example, the independent variable is sex (boys and girls) and it is
nominal; and the dependent variable is test anxiety (level as measured by a test)
and it is interval. The objective of the study is to compare two means and the
samples are independent. Assuming that the assumption of normality and
homogeneity are met, then a parametric test is appropriate.

In this case, we will use independent –samples t-test*.

* The procedure in computing independent samples t-tests and other parametric

tests, as well as the nonparametric tests will be discussed on the succeeding
modules, using SPSS.
Step 4. Establish the critical (rejection) region

Determine the value (and the region) that the test statistics must attain to
be declared statistically significant, that is to be able to reject the null hypothesis.
The level of significance determines which values of the test statistics have very
low probability of occurrence under the null hypothesis. Most often, all absolute
values of the test statistics greater than or equal to these values have very low
probability of occurrence under the null hypothesis. Therefore, if the test statistic
computed for the sample data falls in this region, then there is a basis for rejecting
the null hypothesis.

The critical values (CV) are summarized below.

Two-tailed One-tailed (right) One-tailed (left)

Null Hypothesis (Ho) = ≤ ≥
Alternate Hypothesis (Ha) ≠ > <
α CV α CV α CV
0.10 ± 1.65 0.10 1.28 0.10 1.28
0.05 ±1.96 0.05 1.65 0.05 1.65
0.01 ±2.33 0.01 2.33 0.01 2.33

For the two tailed test, the null hypothesis should be rejected when the test
value is in either of the two critical regions. A one-tailed test is either right-tailed
or left tailed, depending on the direction of the inequality of the alternative
hypothesis.

Step 5. Compute the value of the test statistics from the sample data.

Perform the required calculations of the test statistics (from Step 3).
Compare the computed value of the test statistics with the critical value (from Step
4) to determine if it falls within the critical (rejection) region.
If for example, the computed t-statistics for the independent sample t-test is say,
4.23a, and the critical value is ± 1.96. The test value is greater than the critical
value, so it falls on the rejection region.

a
This is a hypothetical t-value. The procedure of computing this will be
discussed in detail in the succeeding modules.

Step 6. State your conclusion

Reject the null hypothesis if the test statistics has a value that fall into the
critical (rejection) region; otherwise do not reject the null hypothesis. If the null
hypothesis is rejected, then the alternative hypothesis is the conclusion. However,
if there is no sufficient evidence to reject the null hypothesis, it is retained but not
accepted.

In the example, since the test value (t= 4.23) falls in the critical region, there is
enough evidence to reject the null hypothesis stating that there is no difference on
the test anxiety between boys and girls. The conclusion is that boys and girls
experience different levels of test anxiety (the alternative hypothesis).

Illustration 1. In a recent survey, the average salary of sales managers in Metro

Manila is more than P40, 000 per month and the standard deviation is P6, 400 .
A sample of 30 sales managers has a mean salary of P41, 800. At α = 0.05, test
the claim that the sales managers earn more than P40, 000 per month.

Step 1. State the hypothesis and identify the claim

Ho : μ ≤ P40,000 : It is not more than P40,000 (It could be lower or equal)

Ha: μ > P40, 000 (the claim): It is more than P40,00

Step 2. Level of significance is 0.05

Step 3. The appropriate test statistics is a z-test**. ( It is used when n≥ 30, or when
the population is normally distributed and the population standard deviation is
known).

𝑆𝑎𝑚𝑝𝑙𝑒 𝑚𝑒𝑎𝑛−ℎ𝑦𝑝𝑜𝑡ℎ𝑒𝑠𝑖𝑧𝑒𝑑 𝑚𝑒𝑎𝑛

z = 𝑝𝑜𝑝𝑢𝑙𝑎𝑡𝑖𝑜𝑛 𝑠𝑡𝑎𝑛𝑑𝑎𝑟𝑑 𝑑𝑒𝑣𝑖𝑡𝑖𝑜𝑛 ÷
√𝑛

**In the succeeding modules, the Special Package for Social Sciences
(SPSS) software will be utilized, to avoid long calculations.

Step 4. Compute the test value. In this example

41,800−40,000
z= = 1.54
6,400 ÷ √30

Step 5. This is a right-tailed test, the critical value is z= 1.65. The test value (z) is
less than the critical value.

Step 6. Do not reject (retain) the null hypothesis, since the test value falls in the
noncritical region. The conclusion is that the average salary of sales managers is
not more than P40, 000.

Illustration 2. A recruitment agency claims that the average starting salary for call
center agents in Pampanga is P24,000 per month. A sample of 10 call center
agents has a mean salary of P23,220 and a standard deviation of P400. Is there
enough evidence to reject the claim?

Step 1. State the hypotheses and identify the claim

Ho : μ = P 24,000 (the claim) : It is equal to P24,000

Ha: μ ≠ P24,000 : It is not equal to P24,00

Step 2. Level of significance is 0.05

Step 3. The appropriate test statistics is a t-test (It is used when n < 30, or when
the population is normally distributed and the population standard deviation
is unknown).

𝑆𝑎𝑚𝑝𝑙𝑒 𝑚𝑒𝑎𝑛−ℎ𝑦𝑝𝑜𝑡ℎ𝑒𝑠𝑖𝑧𝑒𝑑 𝑚𝑒𝑎𝑛

t= , the degrees of freedom (df) is n-1
𝑠𝑎𝑚𝑝𝑙𝑒 𝑠𝑡𝑎𝑛𝑑𝑎𝑟𝑑 𝑑𝑒𝑣𝑖𝑡𝑖𝑜𝑛 ÷ √𝑛

Step 4. Compute the test value. In this example,

23,220−24,000
t= = - 6.17
400 ÷ √10

Step 5. This is a two tailed-tailed test. Using the table for critical values of the
t-Distribution*** (which can be found in most statistics textbooks as appendices),
the critical value is ±2.262. The test value is greater than the critical value.

Step 6. There is enough evidence to reject the claim that the starting salary of call
center agents in Pampanga is P24, 000. The conclusion is that the average salary
of sales managers is not equal to P24, 000. It is significantly lower.

*** In the succeeding modules where SPSS will be utilized, the table for
critical values of the t-Distribution is not necessary. The software includes
an output that computes the value of the test statistics and compares it
with the critical value at a certain level of significance.
 ACTIVITIES

I. Fill in the blank with a word/s that will make the statement complete and correct.

1. The probability of a Type I error is determined by __________.

2. The probability of a Type II error is determined by _________.

3. If we reject Ho when it is true, we have made a ________ error

4. If we reject a true null hypothesis, we have made a _______ decision.

5. If we fail to reject Ho when it is is false, we have made a ________error

6. If we reject a false null hypothesis, we have made a _________ decision.

7. If we reject the null hypothesis which is false, the conclusion is __________.

8. If the test value falls on the noncritical region, we _____ the null hypothesis.

9. If we have a priori knowledge about the sample value to expect, then we will
use the_______ tests.
10. If the test value is greater than the critical value, we _____ the null hypothesis

II. State the null and alternative hypotheses of the following.

1. The average age of taxi drivers in Makati City is 33 years od.

2. The average income of street sweepers in Quezon City is P6,500.

3. The IQ’s of scientists are greater than 120.

4. The average height of UAAP basketball players is 5’10”.

5. The IQ of doctors and lawyers are the same.

6. The average cost of mobile phones (cell phone) is P10,500.

7. There is no difference in Statistics grades between male and female students.

8. The NAT scores of students in public and private schools are equal.

9. The average phone bill for Valenzuela City residents is greater than P1,500.

10. The average monthly income of private elementary school teachers in Nueva
Ecija is less than P8,000
III. The following problems were to be investigated. Formulate a null hypothesis
and an alternative hypothesis for each of the research problems. Tell whether a
directional or non-directional test is applicable.

1. Do male students feel the same way about long-distance relationships as

female students do?

2. Is pineapple juice more effective in reducing cholesterol levels than lemon

juice?

3. Does the new teaching method affect the performance of students in

Mathematics?

4. Is passing the Licensure Examination for Teachers (LET) dependent on the

type of school graduated from?

5. Do students who group-reviewed better than those who self-reviewed?

IV. . Illustrate the steps of hypothesis testing using the problem below.

Happy Party, a chairs and table rental company claims that the average
number of chairs rented in a party is 250 chairs. A sample of 20 rentals has an
average of 230 chairs with a standard deviation of 20. At α =0.05, is there enough
evidence to reject Happy Party’s claim? Note: Using the critical value of the t –
distribution, the critical value = 2.093 (two-tailed), if df = 19 (n-1).