Topic 6

Two-Variable
Regression: Interval
Estimation and
Hypothesis Testing
UBEQ2013 Basic Econometrics

May 2015/16: Week 6 1

6.1 Introduction
Important concepts:
(i)

probability

(ii) probability distribution

(iii) Type I and Type II errors
(iv) level of significance
(v) power of a statistical test
(vi) confidence interval
UBEQ2013 Basic Econometrics

May 2015/16: Week 6 2

6.1 Introduction
In topic 4, the OLS estimators
are point estimators.
For example, is a point estimate
of unknown population value .
Because of sample fluctuations,
a single estimate is likely to
differ from the true value.

UBEQ2013 Basic Econometrics

May 2015/16: Week 6 3

6.2 Interval Estimation

In statistics, the reliability of a point
estimator is measured by standard error.
Therefore, we may construct an interval
around the point estimator, say within two
or three standard errors on either side of
the point estimator.
Such that this interval has, say, 95%
probability of including the true parameter
value the idea of interval estimation.

UBEQ2013 Basic Econometrics

May 2015/16: Week 6 4

6.2 Interval Estimation

Assume that we want to find out how
close, say, is to .
For this purpose we try to find out
two positive numbers
and , the
)
latter lying between 0 and 1.
The probability that the random
interval contains the true value is .

UBEQ2013 Basic Econometrics

May 2015/16: Week 6 5

6.2 Interval Estimation

Eq. (6.1) is known as a confidence
interval (C.I.).
is known as the confidence coefficient
is known as the level of significance
(probability of committing a Type I
error).
The endpoints of the C.I. are known as
the confidence limits (or critical
values)
UBEQ2013 Basic Econometrics

May 2015/16: Week 6 6

6.2 Interval Estimation

being lower confidence limit
(critical value) and being the
upper confidence limit (critical
value).
Note that and are often
expressed in percentage forms
as and percent.

UBEQ2013 Basic Econometrics

May 2015/16: Week 6 7

6.2 Interval Estimation

Eq. (6.1) shows that an
interval estimator has a
specified probability of
including within its limits the
Example:
true
value
of
the
parameter.
If , or 5%.
Interpretation of eq. (6.1): The probability that the random interval
shown there includes the true is 0.95, or 95 percent
The interval estimator gives a range of values within which the true
may lie.
UBEQ2013 Basic Econometrics

May 2015/16: Week 6 8

6.2 Interval Estimation

(Important Notes)
Eq. (6.1) does not say that the
probability of lying between
the given limits is .
Instead, it states that the
probability of constructing an
interval that contain is .
UBEQ2013 Basic Econometrics

May 2015/16: Week 6 9

6.2 Interval Estimation

(Important Notes)
The interval in Eq. (6.1) is a random
interval.
It will vary from one sample to the next
because it is based on , which is random.
(Its value changes from sample to sample as data change)

Since the C.I. is random, the probability

statements attached to it should be
understood in repeated sampling.
UBEQ2013 Basic Econometrics

May 2015/16: Week 6 10

6.2 Interval Estimation

(Important Notes)
Eq. (6.1) means: If in repeated
sampling C.Is are constructed a great
many times on the basis, then, in the
long run, on the average, such
intervals will enclose in of the cases
the true value of the parameter.
But if is not random or is known, we
cannot make the probabilistic
statement in Eq. (6.1).
UBEQ2013 Basic Econometrics

May 2015/16: Week 6 11

6.3 Confidence Interval for

Regression Coefficients and
With normality assumption for ,
the OLS estimators and are
themselves normally
distributed.
The standardized normal
variable is written as

UBEQ2013 Basic Econometrics

May 2015/16: Week 6 12

6.3 Confidence Interval for

Regression Coefficients and
We use the unbiased estimator
to predict . Next, we replace by
in Eq. (6.2).
(estimator
It becomes
- parameter)

(estimated standard error of estimator)

UBEQ2013 Basic Econometrics

May 2015/16: Week 6 13

6.3 Confidence Interval for

Regression Coefficients and
Instead of using the normal
distribution, we can use the t
distribution to establish a C.I. for :

CI for is written as:

(See Gujarati & Porter (2009) pp 110 for more details on the formulas)
UBEQ2013 Basic Econometrics

May 2015/16: Week 6 14

6.3 Confidence Interval for

Regression Coefficients and
Both Eq. (6.4) and (6.5) shows that:
(i) the larger the se, the larger is the
width
of CI;
(ii) the larger the , the greater is the
uncertainty of estimating the true
value
of the unknown parameter .
UBEQ2013 Basic Econometrics

May 2015/16: Week 6 15

6.4 Hypothesis Testing

The statistical hypothesis testing
allows us to answer the question
of whether a finding is sufficiently
close to the hypothesized value so
that we do not reject the stated
hypothesis (null hypothesis, H0).
H0 is usually is tested against an
alternative hypothesis, H1.
UBEQ2013 Basic Econometrics

May 2015/16: Week 6 16

6.4 Hypothesis Testing

Hypothesis may be simple or
composite.
A simple hypothesis specifies the
precise value(s) of the parameter(s)
of a probability density function.
A composite hypothesis does not
specify the precise value(s) of the
(simple hypothesis)
parameter(s).
(composite hypothesis)
UBEQ2013 Basic Econometrics

May 2015/16: Week 6 17

6.4 Hypothesis Testing

To decide whether to reject or not reject H0 ,
there are two mutually complementary
approaches:
(i) confidence interval (we use either one)
(ii) test of significance
Both of these approaches predicate that the
variable (statistic or parameter) under
consideration has some probability
distribution.
UBEQ2013 Basic Econometrics

May 2015/16: Week 6 18

6.5 Hypothesis Testing: The

Confidence-Interval Approach
Results of a regression analysis (example)
Table 6.1 Regression results

Note: n = 13;
UBEQ2013 Basic Econometrics

May 2015/16: Week 6 19

6.5 Hypothesis Testing: The

Confidence-Interval Approach
Two-Sided or Two-Tail

Test

(composite or two-sided hypothesis)

We postulate that the true slope

coefficient is 0.5 under null hypothesis
but less than or greater than 0.5 under
the alternative hypothesis.
UBEQ2013 Basic Econometrics

May 2015/16: Week 6 20

6.5 Hypothesis Testing: The

Confidence-Interval Approach

Values of lying in this

interval are plausible under
with confidence. Hence,
do not reject if lies in this
region

UBEQ2013 Basic Econometrics

Figure 6.1 A
confidence
interval for

May 2015/16: Week 6 21

6.5 Hypothesis Testing: The

Confidence-Interval Approach
Two-Sided or Two-Tail
Test
H1 is composite (two-sided).

It reflects the fact of lacking a

strong a priori or theoretical
expectation about the direction in
which H1 should move from H0.

UBEQ2013 Basic Econometrics

May 2015/16: Week 6 22

6.5 Hypothesis Testing: The

Confidence-Interval Approach
Is there a significant relationship
between Yi and Xi ?
1.

(Refer to Table 6.1)

2. Critical value:
3. Decision rule: If falls outside this C.I., reject .
Otherwise, do not reject .
(continued next page)
UBEQ2013 Basic Econometrics

May 2015/16: Week 6 23

6.5 Hypothesis Testing: The

Confidence-Interval Approach
4. C.I. :
5. Decision: Reject . The hypothesized value () falls outside the C.I.
6. Conclusion: There is a significant relationship between Yi and Xi.
Xi is statistically significant.

UBEQ2013 Basic Econometrics

May 2015/16: Week 6 24

6.5 Hypothesis Testing: The

Confidence-Interval Approach
One-Sided or One-Tail

Test

(composite or one-sided hypothesis)

We use one-sided test if we have a

strong a priori or theoretical
expectation
UBEQ2013 Basic Econometrics

May 2015/16: Week 6 25

6.6 Hypothesis Testing: The Testof-Significance Approach

Testing the Significance of Regression
Coefficient: The t Test
It is an alternative but complementary
approach to the C.I. method of testing
statistical hypothesis.
It is a procedure by which sample results
are sued to verify the truth or falsity of H0.
Eq.(6.3) is used to compute the value of
the test statistic.

UBEQ2013 Basic Econometrics

May 2015/16: Week 6 26

6.6 Hypothesis Testing: The Testof-Significance Approach

Testing the Significance of Regression
Coefficient: The t Test
Is there a significant relationship
1.
between Yi Twoand Xi ?
(Refer to Table 6.1)
tail

2. Critical value:
3. Decision rule: Reject if the test statistic is greater than the upper
critical value or smaller than the lower critical
value. Otherwise, do not reject .
UBEQ2013 Basic Econometrics

May 2015/16: Week 6 27

6.6 Hypothesis Testing: The Testof-Significance Approach

Testing the Significance of Regression
Coefficient: The t Test
Is there a significant relationship
4. Test
statistic:
between
Yi and Xi ?
5. Decision: Reject . The test statistic (+10.34) is larger than the
upper critical value (+2.201).
6. Conclusion: There is a significant relationship between Yi and Xi.
Xi is statistically significant.
UBEQ2013 Basic Econometrics

May 2015/16: Week 6 28

6.6 Hypothesis Testing: The Testof-Significance Approach

Testing the Significance of Regression
Coefficient: The t Test

Density

Figure 6.2
The 95%
confidence interval
for t (11 df).

Critical
region
2.5%

UBEQ2013 Basic Econometrics

95% region of
acceptance

t = 10.34 lies
in this critical
region

Critical
region
2.5%

May 2015/16: Week 6 29

6.6 Hypothesis Testing: The Testof-Significance Approach

Testing
the Significance of Regression

Coefficient: The t Test

Notice that if , the t statistic will be
zero.
However, as value departs from the
hypothesized value, will be
increasingly larger.
Therefore, a large value will be
evidence against H0.
UBEQ2013 Basic Econometrics

May 2015/16: Week 6 30

6.6 Hypothesis Testing: The Testof-Significance Approach

Testing the Significance of Regression
Coefficient: The t Test
Table 6.2 The t Test of Significance: Decision Rule

Type of
H0: The
Hypothes Null
is
Hypothes
is

H1: The
alternativ
e
Hypothes
is

Decision
Rule:
Reject H0
if

Two-tail
Right-tail
Note:
is the hypothesized numerical value of .
Left-tail
UBEQ2013 Basic Econometrics

May 2015/16: Week 6 31

6.6 Hypothesis Testing: The

Test-of-Significance Approach

Testing the Significance of :

The 2 Test

The test statistic of 2 is:

Eq. (6.6) follows the df of (n-k-1).

UBEQ2013 Basic Econometrics

May 2015/16: Week 6 32

6.6 Hypothesis Testing: The Testof-Significance Approach

Testing the Significance of :
The 2 Test
Table 6.3 A Summary of the 2Test

H0: The Null

Hypothesis

H1: The
alternative
Hypothesis

Decision Rule:
Reject H0 If

Note: is the hypothesized numerical value of .

UBEQ2013 Basic Econometrics

May 2015/16: Week 6 33

6.7 The Exact Level of

Significance: The p Value
The p value is also known as the
observed or exact level of
significance or the exact probability
of committing a Type I error.
More technically, it is the lowest
significance level at which a null
hypothesis can be rejected.
UBEQ2013 Basic Econometrics

May 2015/16: Week 6 34

6.7 The Exact Level of

Significance: The p Value
Is there a significant relationship
1. between Y and X ?
i
i
(Refer to Table 6.1)

2. Level of significance:
3. Decision rule: Reject if the p value is smaller than . Otherwise, do
not reject .
4. p value = 0.000
5. Decision: Reject . The p value is less than .
6. Conclusion: There is a significant relationship between Yi and Xi .
UBEQ2013 Basic Econometrics

May 2015/16: Week 6 35

Thank You!

UBEQ2013 Basic Econometrics

May 2015/16: Week 6 36