3/29/2022

5. Two-variable Linear Regression

Model: Interval Estimation and
Hypothesis Testing

Interval Estimation and Hypothesis Testing

• Two major branches of classical statistics:
• Estimation
• Hypothesis Testing
• The theory of estimation consists of two parts:
• Point estimation
• Interval estimation
• So far we discussed point estimation using the ordinary
least-squares (OLS) method
• Now: Interval estimation followed by hypothesis testing
• Hypothesis testing closely related to interval estimation.

1
 3/29/2022

5.2 INTERVAL ESTIMATION: BASIC IDEAS

Interval Estimation: Basic Ideas

• 𝛽መ2 is the point estimate of 𝛽2 (the slope) which gives the
rate of change of the dependent variable 𝑌 as the
independent/explanatory variable 𝑋 changes
• How reliable is this estimate?
• Due to sampling fluctuations, a single estimate is likely to
differ from the true value,
• Although in repeated sampling its mean value is expected to
be equal to the true value [𝐸 𝛽መ2 = 𝛽2 ].
• In statistics the reliability of a point estimator is measured
by its standard error
• 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 percent probability of
including the true parameter value.
4

2
 3/29/2022

Interval Estimation: Basic Ideas

• We want to find out how “close” is, say, 𝛽መ2 to 𝛽2 .
• Two positive numbers: 𝛿 and 𝛼, where 𝛼 ∈ (0, 1)
• Probability that the random interval (𝛽መ2 − 𝛿, 𝛽መ2 + 𝛿)
contains 𝛽2 is 1 − 𝛼
𝑃(𝛽෠2 − 𝛿 ≤ 𝛽2 ≤ 𝛽෠2 + 𝛿) = 1 − 𝛼
• Such an interval is known as a confidence interval
• 1 − 𝛼 is known as the confidence coefficient
• 𝛼 is known as the level of significance, 0 < 𝛼 < 1
• Standard practice: 𝛼 and 1 − 𝛼 expressed in
percentage terms
--> 100𝛼 and 100 1 − 𝛼 percent

Interval Estimation: Basic Ideas

• The endpoints of the confidence interval are known as
the confidence limits (also known as critical values)
• 𝛽መ2 − 𝛿 : lower confidence limit
• 𝛽መ2 + 𝛿 : upper confidence limit.
The expression
𝑃(𝛽መ2 − 𝛿 ≤ 𝛽2 ≤ 𝛽መ2 + 𝛿) = 1 − 𝛼
is an interval estimator
• Example: if 𝛼 = 0.05, or 5 percent: probability that
the (random) interval shown there includes the true 𝛽2
is 0.95, or 95 percent. The interval estimator thus gives
a range of values within which the true 𝛽2 may lie.

3
 3/29/2022

Interval Estimation: Basic Ideas

Some aspects of interval estimation:
1. Interval estimator does not say that the probability of
𝛽2 lying between the given limits is 1 − 𝛼.
• Since 𝛽2, although an unknown, is assumed to be a fixed
number, either it lies in the interval or it does not.
• The probability of constructing an interval that contains
𝛽2 is 1 − 𝛼.
2. The interval is a random interval; that is, it will vary
from one sample to the next because it is based on 𝛽መ2 ,
which is random.

Interval Estimation: Basic Ideas

3. Since the confidence interval is random, the probability
statements attached to it should be understood in the long-
run sense, i.e. repeated sampling.
• If in repeated sampling confidence intervals like it are
constructed a great many times on the 1 − 𝛼 probability
basis, then, in the long run, on the average, such intervals will
enclose in 1 − 𝛼 of the cases the true value of the
parameter.
4. the interval is random as long as 𝛽መ2 is not known.
• If there is a specific sample and we obtain a specific
numerical value of 𝛽መ2 , the interval is no longer random; it is
fixed.
• In that case, we cannot say that the probability is 1 − 𝛼 that
a given fixed interval includes the true 𝛽2 . 𝛽2 is either in the
fixed interval or outside it. Therefore, the probability is either
1 or 0.

4
 3/29/2022

5.3 CONFIDENCE INTERVALS FOR

REGRESSION COEFFICIENTS 𝛽1 AND 𝛽2

Confidence Interval for 𝛽2

• Normality assumption for 𝑢𝑖
the OLS estimators 𝛽መ1 and 𝛽መ2 are themselves normally
distributed
• we can use the normal distribution to make
probabilistic statements about 𝛽2 provided the true
population variance 𝜎 2 is known.
• If 𝜎 2 is known, then for a normally distributed variable
with mean 𝜇 and variance 𝜎 2 :
• area under the normal curve
• 68% between 𝜇 ± 𝜎
• 95% between the limits 𝜇 ± 2𝜎
• 99.7% between 𝜇 ± 3𝜎

10

5
 3/29/2022

Confidence Interval for 𝛽2

𝛽መ2 − 𝛽2 𝛽መ2 − 𝛽2
𝑍= =
𝑠𝑒(𝛽መ2 ) 𝜎𝛽෡ 2
2
• True value of the variance 𝜎 is rarely known
• In practice we use the estimator 𝜎ො 2
𝛽መ2 − 𝛽2 𝑒𝑠𝑡𝑖𝑚𝑎𝑡𝑜𝑟 − 𝑝𝑎𝑟𝑎𝑚𝑒𝑡𝑒𝑟
𝑡= =
𝑠𝑒(𝛽መ2 ) 𝑒𝑠𝑡𝑖𝑚𝑎𝑡𝑒𝑑 𝑠𝑡𝑑. 𝑒𝑟𝑟𝑜𝑟 𝑜𝑓 𝑒𝑠𝑡𝑖𝑚𝑎𝑡𝑜𝑟
𝛽መ2 − 𝛽2
𝑡=
𝜎ො
• 𝑡 variable defined above follows the 𝑡 distribution with
𝑛 − 2 df.

11

Confidence Interval for 𝛽2

• We use the 𝑡 distribution to establish a confidence
interval for 𝛽2 as follows

• 𝑡𝛼Τ2 is the value of the 𝑡 variable obtained from the 𝑡

distribution for 𝛼/2 level of significance and 𝑛 − 2 df
• Called the critical 𝑡 value at 𝛼/2 level of significance

12

6
 3/29/2022

Confidence Interval for 𝛽2

Rearranging

13

Confidence Interval for 𝛽2

• Above expression gives 100(1 − 𝛼) percent

confidence interval for 𝛽2 , which can also be written as:
𝛽መ2 ± 𝑡𝛼Τ2 𝑠𝑒(𝛽መ2 )
• Similarly:

• 100 1 − 𝛼 % confidence interval for 𝛽1

𝛽መ1 ± 𝑡𝛼Τ2 𝑠𝑒(𝛽መ1 )

14

7
 3/29/2022

Confidence Interval for 𝛽2

• Note: in both cases, the width of the confidence
interval is proportional to the standard error of the
estimator
• Larger the standard error, the larger is the width of the
confidence interval
• Larger the standard error of the estimator, the greater is
the uncertainty of estimating the true value of the
unknown parameter
• This is why the standard error of an estimator is
described as a measure of the precision of the estimator
• i.e., how precisely the estimator measures the true
population value

15

Confidence Interval for 𝛽2

• an earlier example:

• Assuming 𝛼 = 5%,  95% confidence coefficient,

then the 𝑡 table shows that for 8 df the critical 𝑡𝛼Τ2 =
𝑡0.025 = 2.306.

16

8
 3/29/2022

𝛼 = 5%,  95% confidence coefficient, then from 𝑡 table for 8 df:

critical 𝑡𝛼Τ2 = 𝑡0.025 = 2.306. 17

Confidence Interval for 𝛽2

• Substituting in the expression

we get: 0.4268 ≤ 𝛽2 ≤ 0.5914

• This can also be written as:
𝛽መ2 ± 𝑡𝛼Τ2 𝑠𝑒(𝛽መ2 )
0.5091 ± 2.306(0.0357)
0.5091 ± 0.0823
• Interpretation: Given the confidence coefficient of 95%,
in the long run, in 95 out of 100 cases, intervals like
(0.4268, 0.5914) will contain the true 𝛽2 .

18

9
 3/29/2022

Confidence Interval for 𝛽1

• As for 𝛽2 , similarly for 𝛽1 , the confidence interval is:

• we have the values as below:

• From the 𝑡 table we found 𝑡𝛼Τ2 = 𝑡0.025 = 2.306 for

𝛼 = 5%, (95% confidence coefficient), and for 8 df.
19

Confidence Interval for 𝛽1

• So the lower confidence limit is
𝛽መ1 − 𝑡𝛼Τ2 𝑠𝑒 𝛽መ1 = 24.4545 − 2.306 6.4138
= 9.6643
• And the upper confidence limit is:
𝛽መ1 + 𝑡𝛼Τ2 𝑠𝑒 𝛽መ1 = 24.4545 + 2.306 6.4138
= 39.2447
• Hence, the confidence interval for 𝛽1 in the
consumption–income example is:
9.6643 ≤ 𝛽1 ≤ 39.2447

20

10
 3/29/2022

Confidence Interval for 𝛽1

• This can also be written as:
𝛽መ1 ± 𝑡𝛼Τ2 𝑠𝑒 𝛽መ1
24.4545 ± 2.306 6.4138
24.4545 ± 14.7902
• Note: Interpreting this confidence interval
• In the long run, in 95 out of 100 cases intervals like the
ones above will contain the true 𝛽1
• The probability that this particular fixed interval includes
the true 𝛽1 is either 1 or 0.

21

5.4 CONFIDENCE INTERVAL FOR 𝜎 2

22

11
 3/29/2022

Confidence Interval for 𝜎 2

Under the normality assumption of 𝑢𝑖
• 𝑛 − 2 𝜎ො 2 Τ𝜎 2 is distributed as the 𝜒 2 (chi-square)
distribution with 𝑛 − 2 df.
2
𝜎ො 2
𝜒 = 𝑛− 2
𝜎2
• Therefore, we can use the 𝜒 2 distribution to establish a
confidence interval for 𝜎 2 , which is:
2 2 2
𝑃 𝜒1− 𝛼ൗ ≤ 𝜒 ≤ 𝜒𝛼 Τ2 = 1 − 𝛼
2
2
• 𝜒1− 𝛼Τ
2
and 𝜒𝛼2 Τ2
are the critical 𝜒 2 values obtained
from the chi-square table for 𝑛 − 2 df

23

Confidence Interval for 𝜎 2

2 2 2
𝑃 𝜒1−𝛼ൗ ≤ 𝜒 ≤ 𝜒𝛼 Τ2 = 1 − 𝛼
2
2
• Substituting 𝜒 from the earlier expression and rearranging,
we get:

• Example Contd.: 𝜎ො 2 = 42.1591 and df = 8. If 𝛼 = 0.05,

2
• chi-square table for 8 df: critical values: 𝜒0.025 = 17.5346,
2
and 𝜒0.975 = 2.1797.
• These values show that the probability of a chi-square value
exceeding 17.5346 is 2.5 percent and that of 2.1797 is 97.5
percent.
24

12
 3/29/2022

25

• The interval between these two values is the 95%

confidence interval for 𝜒 2 26

13
 3/29/2022

Confidence Interval for 𝜎 2

• Lower confidence limit
𝜎ො 2 42.1591
𝑛− 2 2 =8× = 19.2347
𝜒𝛼Τ2 17.5346
• Upper confidence limit
𝜎ො 2 42.1591
𝑛− 2 2 =8× = 154.7336
𝜒1−𝛼ൗ 2.1797
2
• The 95% confidence interval for 𝜎 2 is:
19.2347 ≤ 𝜎 2 ≤ 154.7336
• Interpretation of this interval: If we establish 95%
confidence limits on 𝜎 2 and maintain from the start
that these limits will include true 𝜎 2 , we shall be right
in the long run 95 percent of the time. 27

5.5 HYPOTHESIS TESTING

28

14
 3/29/2022

Hypothesis Testing: Basic Idea

Statistical Hypothesis Testing
• Is a given observation or finding compatible with some
stated hypothesis or not?
• “compatible,” implies “sufficiently” close to the
hypothesized value
• So that the stated hypothesis is not rejected
• E.g. In the consumption expenditure–income example
• If some theory or past experience made us believe that
true slope coefficient 𝛽2 = 1
• Then, is the observed 𝛽መ2 = 0.5091 obtained from the
sample consistent with the stated hypothesis?
• If it is, we do not reject the hypothesis, else we may
reject it
29

Hypothesis Testing: Basic Idea

• The stated hypothesis is known as the null hypothesis
and is denoted by the symbol 𝐻0
• which may state, e.g. that true 𝛽2 = 1
• The null hypothesis is usually tested against an
alternative hypothesis (also known as maintained
hypothesis) denoted by 𝐻1
• which may state, e.g. that true 𝛽2 ≠ 1
• The alternative hypothesis may be simple or composite
• E.g. simple hypothesis  𝐻1 : 𝛽2 = 1.5
• E.g. composite hypothesis  𝐻1 : 𝛽2 ≠ 1.5

30

15
 3/29/2022

Hypothesis Testing: Basic Idea

• Hypothesis Testing works by developing rules or
procedures for deciding whether to reject or not reject
the null hypothesis.
• There are two mutually complementary approaches for
devising such rules
• confidence interval and
• test of significance.
• Both these approaches predicate that the variable
(statistic or estimator) under consideration has some
probability distribution
• hypothesis testing involves making statements or
assertions about the value(s) of the parameter(s) of
such distribution.

31

Hypothesis Testing: Basic Idea

• E.g. with normality assumption 𝛽መ2 is normally
distributed
𝛽መ2 ~𝑁 𝛽2 , 𝜎𝛽෡2
2
• If we hypothesize that 𝛽2 = 1, we are making an
assertion about one of the parameters of the normal
distribution, namely, the mean.
• Most hypothesis testing we do will be of this type
• Making assertions about one or more values of the
parameters of some assumed probability distribution
• such as the normal, 𝐹, 𝑡, or 𝜒 2

32

16
 3/29/2022

5.6 HYPOTHESIS TESTING: CONFIDENCE-

INTERVAL APPROACH

33

Hypothesis Testing: Confidence-interval

Approach
Two-Sided or Two-Tail Test
• Going back to the example: the estimated marginal
propensity to consume (MPC), 𝛽መ2 = 0.5091
• Say,
𝐻0 : 𝛽2 = 0.3
𝐻1 : 𝛽2 ≠ 0.3
• Null hypothesis: the true MPC (𝛽2 ) is 0.3
• Alternative hypothesis: true MPC (𝛽2 ) is less than/greater
than 0.3
• Composite hypothesis
• a two-sided hypothesis.
• Two-sided alternative hypothesis implies we do not have a
strong expectation about the direction in which the
alternative hypothesis should move from the null
hypothesis.
34

17
 3/29/2022

Hypothesis Testing: Confidence-interval

Approach
Two-Sided or Two-Tail Test
• Is the observed 𝛽መ2 compatible with 𝐻0 ?
• Confidence interval calculated earlier
0.4268 ≤ 𝛽2 ≤ 0.5914
• In the long run intervals like (0.4268, 0.5914) will
contain the true 𝛽2 with 95% probability.
• Thus, the confidence interval provides a set of plausible
null hypotheses.
• So, if 𝛽2 under 𝐻0 falls within the 100(1 − 𝛼)%
confidence interval, we do not reject the null
hypothesis;
• if it lies outside the interval, we may reject it.
35

Hypothesis Testing: Confidence-interval

Approach
Two-Sided or Two-Tail Test
• Decision Rule: Construct a 100(1 − 𝛼)% confidence
interval for 𝛽2 .
• If the 𝛽2 under 𝐻0 falls within this confidence interval,
do not reject 𝐻0 ,
• but if it falls outside this interval, reject 𝐻0 .
• In this example, 𝐻0 : 𝛽2 = 0.3 clearly lies outside the
95% confidence interval. Therefore, we can reject the
hypothesis that the true MPC is 0.3, with 95%
confidence.
• Side Note: If the null hypothesis were true, 𝐻0 : 𝛽2 =
0.3
• The probability of our obtaining a value of MPC of
0.5091 by sheer fluke is at the most about 5%, a small
probability.
36

18
 3/29/2022

Hypothesis Testing: Confidence-interval

Approach
• Confidence interval for 𝛽2:
𝛽መ2 − 𝑡𝛼Τ2 𝑠𝑒 𝛽መ2 , 𝛽መ2 + 𝑡𝛼Τ2 𝑠𝑒 𝛽መ2
𝑃 𝛽መ2 − 𝑡𝛼Τ2 𝑠𝑒 𝛽መ2 ≤ 𝛽2 ≤ 𝛽መ2 + 𝑡𝛼Τ2 𝑠𝑒 𝛽መ2 = 1−𝛼

37

Hypothesis Testing: Confidence-interval

Approach

• When we reject the null hypothesis, we say our

finding(s) is(are) statistically significant.
• When we do not reject the null hypothesis, we say our
finding(s) is(are) not statistically significant.

38

19
 3/29/2022

Hypothesis Testing: Confidence-interval

Approach
One-Sided or One-Tail Test
• When there is a strong expectation (theoretical or
based on previous empirical work) that the alternative
hypothesis is one-sided or unidirectional rather than
two-sided
• Say,
𝐻0 : 𝛽2 ≤ 0.3
𝐻1 : 𝛽2 > 0.3

39

5.7 HYPOTHESIS TESTING: TEST-OF-

SIGNIFICANCE APPROACH

40

20
 3/29/2022

Hypothesis Testing: Test-of-Significance

Approach
Testing the Significance of Regression Coefficients
The 𝒕 Test
• Test-of-significance approach
• An alternative but complementary approach to the
confidence-interval method of testing statistical
hypotheses
• Broadly, a test of significance is a procedure by which
sample results are used to verify the truth or falsity of a
null hypothesis 𝐻0 .

41

Hypothesis Testing: Test-of-Significance

Approach
• under the normality assumption,
𝛽መ2 − 𝛽2
𝑡=
𝑠𝑒(𝛽መ2 )
(𝛽መ2 −𝛽2 ) ∑𝑥𝑖2
𝑡=
𝜎ො
• If the value of true 𝛽2 is specified under the null
hypothesis, the t value can readily be computed from
the available sample, and therefore it can serve as a
test statistic.

42

21
 3/29/2022

Hypothesis Testing: Test-of-Significance

Approach
Since this test statistic follows the 𝑡 distribution, confidence-
interval statements such as the following can be made

where 𝛽2∗ is the value of 𝛽2 under 𝐻0 and where −𝑡𝛼Τ2 and

𝑡𝛼Τ2 are the values of 𝑡 (the critical 𝑡 values) obtained from
the 𝑡 table for (𝛼/2) level of significance and 𝑛 − 2 df
43

Hypothesis Testing: Test-of-Significance

Approach
• Rearranging we get,
𝑃 𝛽2∗ − 𝑡𝛼Τ2 𝑠𝑒 𝛽መ2 ≤ 𝛽መ2 ≤ 𝛽2∗ + 𝑡𝛼Τ2 𝑠𝑒 𝛽መ2 =1−𝛼

• The above expression gives the interval in which 𝛽መ2 will

fall with 1 − 𝛼 probability, given 𝛽2 = 𝛽2∗ .
• The 100(1 − 𝛼)% confidence interval is known as the
region of acceptance (of the null hypothesis)
• The region(s) outside the confidence interval is (are)
called the region(s) of rejection (of 𝐻0 )
• or the critical region(s).

44

22
 3/29/2022

Hypothesis Testing: Test-of-Significance

Approach
• Confidence-interval approach: establish a range or an
interval that has a certain probability of including the
true (but unknown) value of 𝛽2
• Test-of-significance approach: hypothesize some value
for true 𝛽2 then check whether the computed 𝛽መ2 lies
within reasonable (confidence) limits around the
hypothesized value.

45

Hypothesis Testing: Test-of-Significance

Approach
• In the consumption expenditure–income example.
• 𝛽መ2 = 0.5091, 𝑠𝑒 𝛽መ2 = 0.0357, and df = 8.
• If we assume 𝛼 = 5%, then 𝑡𝛼Τ2 = 2.306
• Say,
𝐻0 : 𝛽2 = 𝛽2∗ = 0.3
𝐻1 : 𝛽2 ≠ 0.3
As 𝑡𝛼Τ2 𝑠𝑒 𝛽መ2 = 2.306 0.0357 = 0.0823, then
𝑃 𝛽2∗ − 𝑡𝛼Τ2 𝑠𝑒 𝛽መ2 ≤ 𝛽መ2 ≤ 𝛽2∗ + 𝑡𝛼Τ2 𝑠𝑒 𝛽መ2 =1−𝛼
Becomes
𝑃 0.3 − 0.0823 ≤ 𝛽መ2 ≤ 0.3 + 0.0823 = 1 − 0.05
𝑃 0.2177 ≤ 𝛽መ2 ≤ 0.3823 = 0.95
But observed 𝛽መ2 = 0.5091 is outside this interval 46

23
 3/29/2022

Hypothesis Testing: Test-of-Significance

Approach
• Since the observed 𝛽መ2 = 0.5091 lies in the critical region, we
reject the null hypothesis that true 𝛽2 = 0.3

47

Hypothesis Testing: Test-of-Significance

Approach
• Actually there is no need to construct the interval
explicitly.
• One can just compute the 𝑡 value in the middle of the
double inequality
𝑃 −𝑡𝛼Τ2 ≤ 𝑡 ≤ 𝑡𝛼Τ2 = 1 − 𝛼
𝛽መ2 − 𝛽2∗
𝑡=
𝑠𝑒(𝛽መ2 )
• and check whether it lies between the critical 𝑡 values
or outside them.

48

24
 3/29/2022

Hypothesis Testing: Test-of-Significance

Approach
• In our example, we had found 𝑡𝛼Τ2 = 2.306 for 𝛼 =
0.05 and df = 8
Then
𝑃 −𝑡𝛼Τ2 ≤ 𝑡 ≤ 𝑡𝛼Τ2 = 1 − 𝛼
𝑃 −2.306 ≤ 𝑡 ≤ 2.306 = 1 − 0.05
Since
0.5091 − 0.3
𝑡 = = 5.86
0.0357
• clearly lies in the critical region(outside the interval)
• Conclusion remains the same; we reject 𝐻0 : 𝛽2 = 0.3

49

Hypothesis Testing: Test-of-Significance

Approach
• We reject the null hypothesis that true 𝛽2 = 0.3 since the
observed 𝑡 = 5.86 lies in the critical region

50

25
 3/29/2022

Hypothesis Testing: Test-of-Significance

Approach
• The 𝑡 value will be zero if 𝛽መ2 = 𝛽2∗
• As 𝛽መ2 moves further away (departs) from the
hypothesized 𝛽2 value (𝛽2∗ ), or 𝛽መ2 − 𝛽2∗ increases
• Then |𝑡| goes on increasing in value
• Therefore, a “large” |𝑡| value will be evidence against
the null hypothesis

51

Hypothesis Testing: Test-of-Significance

Approach
Language of reporting significance tests
• A statistic is said to be statistically significant if the
value of the test statistic lies in the critical region.
• The null hypothesis is rejected.
• A test is said to be statistically insignificant if the value
of the test statistic lies in the acceptance region.
• The null hypothesis is not rejected.
• In our example, the 𝑡 test is significant and hence we
reject the null hypothesis.

52

26
 3/29/2022

Hypothesis Testing: Test-of-Significance

Approach
• What we discussed above was a two-sided, or two-tail,
test-of-significance procedure
• We considered the two extreme tails of the relevant
probability distribution
• The rejection regions, and reject the null hypothesis if it
lies in either tail.
• Getting to a One-Sided or One-Tail Test
• Say,
𝐻0 : 𝛽2 ≤ 0.3
𝐻1 : 𝛽2 > 0.3
• this one-tail/sided test will be considering the right tail

53

Hypothesis Testing: Test-of-Significance

Approach
• Test procedure is the same as before
• Except that the upper confidence limit or critical value
now corresponds to 𝑡𝛼 = 𝑡0.05 (5 % significance level)
• We need not consider the lower tail of the 𝑡
distribution in this case.
• Whether one uses a two- or one-tail test of significance
will depend upon how the alternative hypothesis (𝐻1 )
is formulated
• which, itself may depend upon some a priori
considerations or prior empirical experience.

54

27
 3/29/2022

𝑡0.05 = 1.860 for df = 8 55

56

28
 3/29/2022

Hypothesis Testing: Test-of-Significance

Approach

57

Hypothesis Testing: Test-of-Significance

Approach
Testing the Significance of 𝝈𝟐
The 𝝌𝟐 Test
𝜎ො 2
𝑛− 2 ~𝜒 2
𝜎2
𝜎ො 2
𝜒2 = 𝑛 − 2
𝜎2
• Follows the 𝜒 2 distribution with 𝑛 − 2 degrees of
freedom
• In the hypothetical example: 𝜎ො 2 = 42.1591 and df = 8.
• If we postulate that
𝐻0 : 𝜎 2 = 85
𝐻1 : 𝜎 2 ≠ 85 58

29
 3/29/2022

Hypothesis Testing: Test-of-Significance

Approach: The 𝝌𝟐 Test
• Substituting the values in the above equation:
2
𝜎ො 2
𝜒 = 𝑛− 2
𝜎2
42.16
𝜒2 = 8 = 3.97
85
• assuming 𝛼 = 5%, the critical 𝜒 2 values are:
2 2
𝜒1− 𝛼Τ and 𝜒𝛼Τ
2 2

• from:
2 2 2
𝑃 𝜒1− 𝛼ൗ ≤ 𝜒 ≤ 𝜒𝛼 Τ2 = 1 − 𝛼
2

59

60

30
 3/29/2022

Hypothesis Testing: Test-of-Significance

Approach

2 2
• 𝜒0.025 = 17.5346, and 𝜒0.975 = 2.1797.
• Since the computed 𝜒 2 lies between these limits
• The data support the null hypothesis and we do not
reject it.
• This test procedure is called the chi-square test of
significance.

61

Hypothesis Testing: Test-of-Significance

Approach: The 𝝌𝟐 Test

62

31
 3/29/2022

5.8 HYPOTHESIS TESTING: SOME PRACTICAL

ASPECTS

63

Hypothesis Testing: Some Practical Aspects

Meaning of “Accepting” or “Rejecting” a Hypothesis
• If on the basis of a test of significance, say, the 𝑡 test,
we decide to “accept” the null hypothesis,
• All we are saying is that on the basis of the sample
evidence we have no reason to reject it
• we are not saying that the null hypothesis is true beyond
any doubt.
• in “accepting” a null hypothesis we should always be
aware that another null hypothesis may be equally
compatible with the data.
• It is therefore preferable to say that we may accept the
null hypothesis rather than we (do) accept it.
• Just as a court pronounces a verdict as “not guilty”
rather than “innocent,” so the conclusion of a statistical
test is “do not reject” rather than “accept.
64

32
 3/29/2022

Hypothesis Testing: Some Practical Aspects

The “Zero” Null Hypothesis and the “2−𝒕” Rule of

Thumb
• the objective being to find out whether Y is related at
all to X, the explanatory variable.
• If there is no relationship between Y and X to begin
with, then testing a hypothesis such as 𝛽2 = 0.3 or any
other value is meaningless.
• formal testing can be shortcut by adopting the “2-t”
rule of significance

65

Hypothesis Testing: Some Practical Aspects

• “2-𝒕” Rule of Thumb. If the number of degrees of

freedom is 20 or more and if 𝛼 (level of significance), is
set at 0.05, then the null hypothesis 𝛽2 = 0 can be
෡2 −𝛽2∗
𝛽
rejected if the calculated 𝑡 value = ෡2 exceeds 2
𝑠𝑒 𝛽
in absolute value.
• Why this ‘rule’?
• We know that for 𝐻0 : 𝛽2 = 0, if the calculated 𝑡 value
෡ −𝛽∗
𝛽
= 𝑠𝑒2 𝛽෡ 2 does not belong to the interval
2
−𝑡𝛼Τ2 ≤ 𝑡 ≤ 𝑡𝛼Τ2
then, the 𝐻0 will be rejected.

66

33
 3/29/2022

Hypothesis Testing: Some Practical Aspects

• To put it differently, we will reject 𝐻0 if the calculated 𝑡
value is greater than the upper confidence limit
𝑡 > 𝑡𝛼Τ2
𝛽መ2 − 𝛽2∗
 𝑡= > 𝑡𝛼ൗ2
𝑠𝑒 𝛽መ2
Since 𝛽2∗ is the hypothesized value of 𝛽2 , here 𝐻0 : 𝛽2 =
𝛽2∗ = 0, then
𝛽መ2
> 𝑡𝛼ൗ2 for 𝛽መ2 > 0
𝑠𝑒 𝛽መ2
• Similarly, reject 𝐻0 if the calculated 𝑡 value is less than
the lower confidence limit
𝛽መ2
< −𝑡𝛼ൗ2 for 𝛽መ2 < 0
𝑠𝑒 𝛽መ2
67

Hypothesis Testing: Some Practical Aspects

• Combining the two conditions into one,

𝛽෠2
𝑡 = > 𝑡𝛼ൗ
𝑠𝑒 𝛽෠2 2

• On a 𝑡 table, for degree of freedom (df) of about 20 or

more, 𝑡 value in excess of 2 (in absolute terms), is
statistically significant at the 5 percent level
• implying rejection of the null hypothesis.
• So, for df ≥ 20, we do not even have to refer to the 𝑡
table to assess the significance of the estimated slope
coefficient.

68

34
 3/29/2022

Hypothesis Testing: Some Practical Aspects

• For one sided hypothesis 𝛽2 = 0 versus 𝛽2 > 0 or 𝛽2 <

0, then we should reject the null hypothesis if
𝛽෠2
𝑡 = > 𝑡𝛼
𝑠𝑒 𝛽෠2
• If we fix 𝛼 = 0.05, then from the 𝑡 table for 20 or more
df, 𝑡 > 1.73 is statistically significant at the 5% level of
significance (one-tail)

69

Hypothesis Testing: Some Practical Aspects

Forming the Null 𝑯𝟎 and Alternative 𝑯𝟏 Hypotheses
• How to formulate these hypotheses?
• No hard-and-fast rules
• often the phenomenon under study will suggest the nature of
the null and alternative hypotheses
• Example: capital market line (CML) of portfolio theory
𝐸𝑖 = 𝛽1 + 𝛽2 𝜎𝑖
• 𝐸 = expected return on portfolio and 𝜎 = the standard
deviation of return, a measure of risk.
• Since return and risk are expected to be positively related—
the higher the risk, the higher the return
• Then if null hypothesis 𝐻0 : 𝛽2 = 0,
• Naturally the alternative hypothesis would be 𝐻1 : 𝛽2 > 0.
• No point in considering values of 𝛽2 less than zero
70

35
 3/29/2022

Hypothesis Testing: Some Practical Aspects

• Example: Demand for money

• Income - one of the important determinants of the
demand for money
• studies of the money demand functions show that the
income elasticity of demand for money (%∆ in demand
for money for a 1% ∆ in income): 0.7 - 1.3
• So, for a new study of demand for money,
• if one postulates that the income-elasticity coefficient
𝛽2 is 1, or the null hypothesis 𝐻0 : 𝛽2 = 1,
• the alternative hypothesis could be 𝐻1 : 𝛽2 ≠ 1,
• two-sided alternative hypothesis (why?)

71

Hypothesis Testing: Some Practical Aspects

Choosing 𝜶, the Level of Significance

• whether we reject or do not reject the null hypothesis
depends critically on 𝛼: the level of significance
• 𝛼 also the probability of committing a Type I error —
probability of rejecting the true hypothesis
• Type II error: the probability of accepting the false
hypothesis
• if we try to reduce a Type I error, a Type II error
increases, and vice versa.
• So we need to find out the relative costs of the two types
of errors to decide the tradeoff
• Problem: we rarely know the costs of making the two
types of errors

72

36
 3/29/2022

Hypothesis Testing: Some Practical Aspects

• Applied econometricians generally follow the practice

of setting the value of 𝛼 at a 1% or 5% or at most 10%
level and choose a test statistic that would make the
probability of committing a Type II error as small as
possible
• Since one minus the probability of committing a Type II
error is known as the power of the test
• this way we maximizing the power of the test.

73

Hypothesis Testing: Some Practical Aspects

The Exact Level of Significance: The 𝒑 Value

• Problem choosing the appropriate value of 𝛼 can be
avoided if we use what is known as the 𝑝 value of the
test statistic
• Actual probability of obtaining a value of the test
statistic (e.g., the 𝑡 statistic) as much as or greater than
that obtained. This probability is called the 𝑝 value
(probability value)
• 𝑝 value: aka exact level of significance or the exact
probability of committing a Type I error
• 𝑝 value: is defined as the lowest significance level at
which a null hypothesis can be rejected.

74

37
 3/29/2022

Hypothesis Testing: Some Practical Aspects

• Example: consumption–income example
𝐻0 : 𝛽2 = 0.3
The calculated 𝑡 value:
𝛽መ2 − 𝛽2∗ 0.5091 − 0.3
𝑡= = = 5.86
𝑠𝑒(𝛽መ2 ) 0.0357
• What is the 𝑝 value of obtaining a 𝑡 value of as much as
or greater than 5.86?
• 𝑡 table for 8 df the probability of obtaining such a 𝑡
value must be much smaller than even 0.001 (one-tail)
or 0.002 (two-tail)
• using computer: the probability of obtaining a t value of
5.86 or greater (for 8 df) is about 0.000189

75

76

38
 3/29/2022

Hypothesis Testing: Some Practical Aspects

• This observed, or exact, level of significance of the 𝑡

statistic is much smaller than the arbitrarily fixed level
of significance, such as 1, 5, or 10%
• if we were to use the 𝑝 value just computed and reject
the null hypothesis that the true MPC is 0.3, the
probability of our committing a Type I error is only
about 0.02%, that is, only about 2 in 10,000!
• For a given sample size, as |𝑡| increases, the 𝑝 value
decreases
• So one can reject the null hypothesis with increasing
confidence
• It is better to give up fixing 𝛼 arbitrarily at some level
and simply choose the 𝑝 value of the test statistic
77

Hypothesis Testing: Some Practical Aspects

Statistical Significance versus Practical Significance

• In the consumption–income example, if the null
hypothesis said that the true MPC is 0.61
𝐻0 : 𝛽2 = 0.61
• Recall that on the basis of the sample result, 𝛽መ2 =
0.5091, we obtained the interval (0.4268, 0.5914) with
95% confidence.
• Since this interval does not include 0.61, we can, with
95% confidence, say that our estimate is statistically
significant, that is, significantly different from 0.61
• Reminder: statistically significant if the value of the test
statistic lies in the critical region.
• The null hypothesis is rejected.
78

39
 3/29/2022

Hypothesis Testing: Some Practical Aspects

• Practically speaking, what difference does it make if we

take the MPC to be 0.61 rather than 0.5091?
• Is the 0.1009 difference between the two MPCs that
important practically?
• Answer depends on what we really do with these
estimates.
• Macroeconomics:
1
𝑖𝑛𝑐𝑜𝑚𝑒 𝑚𝑢𝑙𝑡𝑖𝑝𝑙𝑖𝑒𝑟 =
1 − 𝑀𝑃𝐶
• if MPC is 0.5091, the multiplier is 2.04, if MPC is equal
to 0.61, then 2.56

79

Hypothesis Testing: Some Practical Aspects

• If the government were to increase its expenditure by

₹1 to lift the economy out of a recession,
• income will eventually increase by ₹2.04 if the MPC is
0.5091
• And by ₹2.56 if the MPC is 0.61
• 25% difference could be very crucial to resuscitating the
economy.
• one should not confuse statistical significance with
practical, or economic, significance.
• As sample size becomes very large, issues of statistical
significance become much less important but issues of
economic significance become critical

80

40
 3/29/2022

5.11 REPORTING THE RESULTS OF

REGRESSION ANALYSIS

81

Reporting the Results of Regression Analysis

• 1st set of parentheses: estimated standard errors of the

regression coefficients
• 2nd set of parentheses: calculated 𝑡 values under the null
hypothesis that true population value of each regression
coefficient individually is zero
• 3rd set of parentheses: estimated 𝑝 values. So, for 8 df the
probability of obtaining a 𝑡 value ≥ 3.8128 is 0.0026 and the
probability of obtaining a 𝑡 value ≥ 14.2605 is 0.0000003. 82

41
 3/29/2022

Reporting the Results of Regression Analysis

• 𝑝 values of the estimated 𝑡 coefficients show

• The exact level of significance of each estimated 𝑡 value.
• Thus, under the null hypothesis that the true population
intercept value is zero, the exact probability (i.e., the 𝑝
value) of obtaining a 𝑡 value of 3.8128 or greater is only
about 0.0026.
• Therefore, if we reject this null hypothesis, the
probability of our committing a Type I error is about 26
in 10,000

83

42