CHAPTER 3

MULTIPLE REGRESSION
MODEL

PREPARED BY:
DR. SITI MULIANA SAMSI
Hypothesis Testing
 Hypothesis testing involves two contrasting propositions (each
called a hypothesis) relating to the value of one or more population
parameters.
 H0: Null hypothesis: describes an existing theory
 H1: Alternative hypothesis: the complement of H0
 Using sample data, we either:
- reject H0 and conclude the sample data provides
sufficient evidence to support H1, or
- fail to reject H0 and conclude the sample data
does not support H1.
 Example: In the U.S. legal system, a defendant is innocent until proven guilty.
H0: Innocent
H1: Guilty
 If evidence (sample data) strongly indicates the defendant is guilty, then we reject
H0.
 Three types of one sample hypothesis tests:
1. H0: parameter ≤ constant
H1: parameter > constant

2. H0: parameter ≥ constant

H1: parameter < constant

3. H0: parameter = constant

H1: parameter ≠ constant
Hypothesis Testing Procedure
Steps in conducting a hypothesis test:
1. Identify the population parameter and formulate the hypotheses to
test.
2. Select a level of significance (the risk of drawing an incorrect
conclusion).
3. Calculate a test statistic.
4. Determine the decision rule on which to base a conclusion.
5. Apply the decision rule and draw a conclusion.
The t-Statistic
  Multiple regression equation:
 We can calculate t-values for each of the estimated coefficients in the equation.
 T-tests are usually done only on the slope coefficients; the relevant form of the t-statistic
for the k-th coefficient is:
Where:
= the estimated regression coefficient of the kth variable
= the border value (usually zero) implied by the null hypothesis for βk
= the estimated standard error of
   
Since most regression hypotheses test whether a particular regression coefficient is
significantly different from zero, βH0 is typically zero. Zero is particularly meaningful
because if the true β equals zero, then the variable doesn’t belong in the equation. Before
we drop the variable from the equation and effectively force the coefficient to be zero,
however, we need to be careful and test the null hypothesis that β=0. Thus, the most-used
form of the t-statistic becomes:
(2.14)
Which simplifies to : (2.15)
 For an example of this calculation, let’s consider the following equation:
(2.16)

standard error (SE)

t-values calculated
according to
Equation 2.15

Number of Obs
  
Using the regression results in Equation 2.16, let’s calculate the t-value for the estimated
coefficient of P, the population variable.
 Given the values in Equation 2.16 of 0.3547 for and 0.0727 for SE(), and given H 0:βP ≤ 0,
the relevant t-value is indeed 4.88, as specified in Equation 2.16:
The Critical t-Value and the t-Test Decision Rule

 To decide whether to reject or not to reject a null hypothesis based on a calculated t-value,
we use a critical t-value.
 A critical t-value is the value that distinguishes the “acceptance” region from the rejection
region.
 The critical t-value, tc, is selected from a t-table (see Statistical Table B-1 on pg. 519 E-
Book: Studenmund 2016) depending on whether the test is one-sided or two-sided, which
we have defined as the number of observations (N) minus the number of coefficients
estimated, which is K (including the constant): N - K - 1.
 Once we have obtained a calculated t-value tk and a critical t-value tc, we can reject the null
hypothesis if the calculated t-value is greater than the critical t-value.
t-value > critical value = Reject H0 and accept HA
t-value < critical value = Do not reject H0 (means that we accept H0)
 Statistical Table B-1 contains the critical values t c for varying degrees of freedom and
levels of significance. The columns indicate the levels of significance according to whether
the test is one-sided or two-sided, and the rows indicate the degrees of freedom.
 Recall that we hypothesized that population’s coefficient would be positive, so this is a one-
sided test:
H0: βp≤0
HA: βp>0

 There are 29 degrees of freedom (equal to N - K - 1, or 33 - 3 - 1) in this regression.

 To find this value, pick a level of significance, say 5 percent, and turn to Statistical Table B-
1.
 The decision rule is to reject H0 if |tk|>tc (t-stat > critical value). Since the 5%, one-sided, 29
degrees of freedom critical t-value is 1.699, and since the sign implied by HA is positive, the
decision rule becomes:
Reject H0 if |tp| >1.699 and if tp is positive
 Since the t-statistics of tp 4.88 is larger than the critical value of 1.699, we would reject the
null hypothesis and conclude that population does indeed tend to have a positive relationship
with Woody’s check volume (holding the other variables in the equation constant).
Choosing a Level of Significance
 The level of significance indicates the probability of observing an estimated t-value greater
than the critical t-value if the null hypothesis were correct.
 How should you choose a level of significance? Most beginning econometricians assume
that the lower the level of significance, the better (5-percent level).
 If we can reject a null hypothesis at the 5-percent level of significance, we can summarize
our results by saying that the coefficient is “statistically significant” at the 5-percent level.
 Thus a t-test for which we use a 5-percent level of significance can also be said to have a
95-percent level of confidence.
 Some researchers produce tables of regression results, and then mark “significant”
coefficients with asterisks.
 10% (*), 5% (**)and 1% (***)
Confidence interval approach
   confidence interval is a range of values that will contain the true value of β a certain
A
percentage of the time, say 90 or 95 percent. The formula for a confidence interval is:

 where tc is the two-sided critical value of the t-statistic for whatever significance level we
choose.
 If you want a 90-percent confidence interval you’d choose the critical value for the 10-
percent significance level. For a 95-percent confidence interval, you’d use a 5-percent
critical value.
 To see how confidence intervals can be used for hypothesis tests, let’s test the significance
of the income coefficient of the Equation below:
  Find 90% confidence interval.  Step 4: Find upper and lower limit
 Step1: Find the t-value 1.288 - 0.923 = 0.365 (lower limit)
100%-90% = 10% 1.288 + 0.923 = 2.21 (upper limit)
= 0.10 (one-tail)  Step 5: Conclusion
= 0.10/2 Since the coefficient of income lies
within the acceptance region of 0.365 ≤ 1.288
= 0.05 (two-tail)
≤ 2.21, based on this interval, we can
df = N-K-1= 33-3-1 = 29  conclude that the variable of income is
tc value, 29,0.05= 1.699 statistically significant to the economic growth
(Y).
(refer t-distribution table)
 Step 2: Calculate Standard Error (SE)
= 0.543
 Step 3: Tc value x SE
1.699 x 0.543 = 0.923
Exercise 3.1
(i) Using the standard normal approximation, find the 95% confidence interval for
βhsGPA.
  Step 1: Find the t-value  Step 5: Conclusion
100% - 95% = 5% or 0.05 (one-tailed) Since the coefficient of βhsGPA lies within the
acceptance region of 0.228 ≤ 0.412 ≤ 0.596, based
= 0.05/2 = 0.025 (two-tailed)
on this interval, we can conclude that the variable
Df = N-K-1 = 141- 3 -1 = 137 of βhsGPA is statistically significant to the
T-critical value,137,0.025 = 1.960 colGPA.

 Step 2: Calculate Standard Error (SE)

= 0.094
 Step 3: t-critical value x SE
1.960 x 0.094 = 0.18424
 Step 4: Find upper & lower limit
Lower limit: 0.412 – 0.18424 = 0.22776
Upper Limit: 0.412 + 0.18242 = 0.59624
(ii) Can you reject the hypothesis H0: βhsGPA = 0.4 against the two-sided alternative
at the 5% level?

  Step 1: State the null and alternative hypothesis  Step 4: Determine the decision rule
H0: βhsGPA = 0.4 Since the t-statistics of 0.1276 is smaller than t-
critical value of 1.960, it falls under the acceptance
H1: βhsGPA ≠ 0.4
region. Therefore, we do not reject H0.
 Step 2: Select the significance level  Step 5: Conclusion
Significance level: 5% or 0.05 (one-tailed)
As the result is fail to reject Ho, we can conclude
= 0.05/2 = 0.025 (two-tailed) that there is sufficient evidence to conclude that
Df = N-K-1 = 141- 3 -1 = 137 βhsGPA is statistically significant equal to 0.4.

T-critical value,137,0.025 = 1.960

 Step 3: Select the t-statistics
 (iii) Can you reject the hypothesis H0: βhsGPA=1 against the two-sided alternative at
the 5% level?

 Step 4: Determine the decision rule

  Step 1: State the null and alternative hypothesis
Since the t-statistics of -6.255 is smaller than t-
H0: βhsGPA = 1
critical value of -1.960, it falls under the rejection
H1: βhsGPA ≠ 1 region. Therefore, we reject H0.
 Step 2: Select the significance level  Step 5: Conclusion
Significance level: 5% or 0.05 (one-tailed) As the result is rejecting Ho, we can conclude that
= 0.05/2 = 0.025 (two-tailed) there is sufficient evidence to conclude that βhsGPA
is statistically significant different from 1.
Df = N-K-1 = 141- 3 -1 = 137
T-critical value,137,0.025 = 1.960 = -1.960
 Step 3: Select the t-statistics
Assumption underlying OLS
The Classical Assumptions

1. The regression model is linear, is correctly specified, and has an

additive error term.
2. The error term has a zero population mean.
3. All explanatory variables are uncorrelated with the error term.
4. Observations of the error term are uncorrelated with each other (no serial
correlation).
5. The error term has a constant variance (no heteroskedasticity).
6. No explanatory variable is a perfect linear function of any other
explanatory variable(s) (no perfect multicollinearity).
7. The error term is normally distributed.
The properties of OLS estimators
Given
  all seven classical assumptions, the OLS coefficient estimators can be shown to have
the following properties:
 1. They are unbiased. That is, is β. This means that the OLS estimates of the coefficients
are centered around the true population values of the parameters being estimated.
 2. They are minimum variance. The distribution of the coefficient estimates around the true
parameter values is as tightly or narrowly distributed as is possible for an unbiased
distribution. No other unbiased estimator has a lower variance for each estimated
coefficient than OLS.
 3. They are consistent. As the sample size approaches infinity, the estimates converge to
the true population parameters. Put differently, as the sample size gets larger, the variance
gets smaller, and each estimate approaches the true value of the coefficient being
estimated.8
 4. They are normally distributed. The are ]). Thus various statistical tests based on the
normal distribution may indeed be applied to these estimates
Estimation of parameters using E-views
statistical software
 Step 1: Select data from excel file of ASEAN-5
including label (y, m, r, p, e) and Copy
 Step 2: Open E-views file –> new -> Workfile
 Step 3: click Quick –> Empty Group  Step 4: Right click –> select “Paste”
 Step 5: generate new log variable (ly, lm, le, lp) excluding r variable.
 Step 6: Estimate the equation of Y=B0+B1M + B2R + B3P + B4E + μ by “CTRL + select
variable” according the model –> click OK
 Step 7: Ordinary Least Square (OLS) Estimation

𝑌  =−1.89+0.41 𝑀 − 0.01 𝑅 +1.72 𝑃 − 0.20 𝐸+𝜖