0% found this document useful (0 votes)
87 views161 pages

Econometrics Notes of Book

The document discusses econometrics and provides examples of its methodology. It defines econometrics as the empirical verification of economic theories using statistics and mathematics. It explains why econometrics is a separate discipline that blends economics, statistics, and mathematics. It then outlines the seven steps of the econometrics methodology, applying it to the consumption function theory.
Copyright
© © All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as PDF, TXT or read online on Scribd
0% found this document useful (0 votes)
87 views161 pages

Econometrics Notes of Book

The document discusses econometrics and provides examples of its methodology. It defines econometrics as the empirical verification of economic theories using statistics and mathematics. It explains why econometrics is a separate discipline that blends economics, statistics, and mathematics. It then outlines the seven steps of the econometrics methodology, applying it to the consumption function theory.
Copyright
© © All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as PDF, TXT or read online on Scribd
You are on page 1/ 161

TOPICS STUDY NOTES

Date : 6th, Mar, 2023

INTRODUCTION

Definition ● Literally interpreted, econometrics means “economic measurement.”


It is a measure of economics of financial theories.
● Empirical Verification of economic laws, i.e Financial theories.
Example: Law of Demand

● Measurement of Theories.
● CAPM Model : Relation between risk and return Q^d = a - bP
● Econometrics is a unique blend of Economics, Statistics and
Mathematics

WHY A As the preceding definitions suggest, econometrics is an amalgam of


SEPARATE economic theory, mathematical economics, economic statistics, and
DISCIPLINE? mathematical statistics.

Example: Theory of consumption.

Economics Theory:
Consumption increases with income.
Statistical working:
Work on consumption (C) , Income (Y) data.

Mathematics:
If Y increase by 1 unit, C increase by C2 units.
C = C1 + C2 Y

C1 = Autonomous consumption that does not depend on Y


C2 = MPC
Slope:

MPC = ΔC / ΔY
MPC + MPS = 1

MPC = Marginal propensity to consume (Always < 1, since we don’t utilize


all increased income)
MPS = Marginal propensity to save

Econometrics:
C = C1 + C2 Y + E
E = Error (C - Ĉ)

Now here, The empirical verification of economic law has been proved.

Date : 13th, Mar, 2023

METHODOLOGY 1. Statement of theory or hypothesis.


OF 2. Specification of the mathematical model of the theory
ECONOMETRICS 3. Specification of the statistical, or econometric, model
4. Obtaining the data
5. Estimation of the parameters of the econometric model
6. Hypothesis testing
7. Forecasting or prediction
8. Using the model for control or policy purposes.
Statement of Keynes’s Theory:
Theory or Keynes stated:
Hypothesis “The fundamental psychological law . . . is that men [women] are disposed,
as a rule and on average, to increase their consumption as their income
increases, but not as much as the increase in their income.”

C and Y are positively related.


0 < MPC < 1

Specification of the C = β1 + β2Y


Mathematical 0 < β2 < 1
Model of β1 and β2, known as the parameters of the model, are, respectively, the
Consumption intercept and slope coefficients.

Specification of the C = β1 + β2Y + u


Econometric u = Error (C - Ĉ)
Model of where u, known as the disturbance, or error, term, is a random (stochastic)
Consumption variable.
We estimate β1 and β2 through OLS
Obtaining Data

Estimation of the We estimate parameters of econometric model β1 and β2 . We estimate β1


Econometric and β2 through OLS
Model C = 0.2 + 0.8Y
β1 = 0.2 (SAVE)
β2 = 0.8 (CONSUME)

Hence proved : 0 < β2 < 1

Hypothesis Testing Ho = MPC should be greater than 0 and less than 1


H1 = MPC should not be greater than 0 and less than 1

Forecasting or If the chosen model does not refute the hypothesis or theory under
Prediction consideration, we may use it to predict the future value(s) of the
dependent, or forecast, variable Y on the basis of known or expected
future value(s) of the explanatory, or predictor, variable X.

Use of the Model An estimated model may be used for control, or policy, purposes. By
for Control or appropriate fiscal and monetary policy mix, the government can
Policy Purposes manipulate the control variable X to produce the desired level of the target
variable Y.
TYPES OF
ECONOMETRICS

CHAPTER # 1 THE NATURE OF REGRESSION ANALYSIS

Date : 15th, Mar, 2023

Single Equation There must be 2 variables. X is independent variable while Y is dependent.


Regression There must be uni-directional causality. (Y is causing X or vice versa).
Model

Origin The term regression was introduced by Francis Galton in 1886.

THE MODERN “Regression analysis is concerned with the study of the dependence
INTERPRETATIO of one variable, the dependent variable, on one or more other
N OF variables, the explanatory variables, with a view to estimating and/or
REGRESSION predicting the (population) mean or average value of the former in
terms of the known or fixed (in repeated sampling) values of the
latter”
regression line passing through averages

STATISTICAL STATISTICAL RELATIONSHIPS:


VERSUS Crop yield = f (temperature, rainfall, sunshine, fertilizer)
DETERMINISTIC u = Tells the impact of variables not considered
RELATIONSHIPS This is a statistical relationship

DETERMINISTIC RELATIONSHIPS:
One to one relation
No error term involved
Solid reasoning and justification
Example: Scientific laws i.e Charles Law, Einstein’s law of energy.

REGRESSION Although regression analysis deals with the dependence of one variable on
VERSUS other variables, it does not necessarily imply causation. In the words of
CAUSATION Kendall and Stuart, “A statistical relationship, however strong and however
suggestive, can never establish causal connection: our ideas of causation
must come from outside statistics, ultimately from some theory or other.”5
In the crop-yield example cited previously, there is no statistical reason to
assume that rainfall does not depend on crop yield. The fact that we treat
crop yield as dependent on rainfall (among other things) is due to
nonstatistical considerations: Common sense suggests that the
relationship cannot be reversed, for we cannot control rainfall by varying
crop yield.

Regression without causation is nothing

REGRESSION Closely related to but conceptually very much different from regression
VERSUS analysis is correlation analysis, where the primary objective is to measure
CORRELATION the strength or degree of linear association between two variables.
Date : 20th, Mar, 2023

TYPES OF DATA

Time Series Data A time series is a set of observations on the values that a variable takes at
different times. Such data may be collected at regular time intervals, such
as daily (e.g., stock prices, weather reports), weekly (e.g., money supply
figures), monthly [e.g., the unemployment rate, the Consumer Price Index
(CPI)], quarterly (e.g., GDP), annually (e.g.,government budgets),
quinquennially, that is, every 5 years (e.g., the census of manufactures), or
decennially (e.g., the census of population).

Cross-Section Cross-section data are data on one or more variables collected at the
Data same point in time, such as the census of population conducted by the
Census Bureau every 10 years (the latest being in year 2000).

Pooled Data In pooled, or combined, data are elements of both time series and
cross-section data.

Panel, Longitudinal, or Micropanel Data:

This is a special type of pooled data in which the same cross-sectional unit
(say, a family or a firm) is surveyed over time. For example, the U.S.
Department of Commerce carries out a census of housing at periodic
intervals.
MEASUREMENT The variables that we will generally encounter fall into four broad
SCALES OF categories:
VARIABLES1

Ratio Scale For a variable X, taking two values, X1 and X2, the ratio X1/X2 and the
distance (X2 − X1) are meaningful quantities. Thus, it is meaningful to ask
how big is this year’s GDP compared with the previous year’s GDP.

Interval Scale An interval scale variable satisfies the last two properties of the ratio scale
variable but not the first. Thus, the distance between two time periods, say
(2000–1995) is meaningful, but not the ratio of two time periods
(2000/1995).

Ordinal Scale A variable belongs to this category only if it satisfies the third property of
the ratio scale (i.e., natural ordering). Examples are grading systems (A, B,
C grades)

Nominal Scale Variables in this category have none of the features of the ratio scale
variables. Variables such as gender (male, female) and marital status
(married, unmarried, divorced, separated) simply denote categories.

EXERCISES

1.1 Table 1.2 gives data on the Consumer Price Index (CPI) for seven
industrialized countries with 1982–1984 = 100 as the base of the index. a.
From the given data, compute the inflation rate for each country.16 b. Plot
the inflation rate for each country against time (i.e., use the horizontal axis
for time and the vertical axis for the inflation rate.) c. What broad
conclusions can you draw about the inflation experience in the seven
countries? d. Which country’s inflation rate seems to be most variable?
Can you offer any explanation?

Answer
1.2

Answer

1.3
Answer

1.4
Answer

1.5

Answer

1.6
Answer

1.7

Answer
CHAPTER # 2 TWO-VARIABLE REGRESSION ANALYSIS: SOME BASIC
IDEAS

Conditional mean Mean at a given condition, Example: Income at a given consumption level.
Regression line passes through conditional means.

Example:
Date : 22th, Mar, 2023

population If we join these conditional mean values, we obtain what is known as the
regression line population regression line (PRL), or more generally, the population
(PRL), regression curve.5 More simply, it is the regression of Y on X.

POPULATION conditional mean E(Y | Xi) is a function of Xi, where Xi is a given value of X.
REGRESSION Symbolically,
FUNCTION (PRF)
E(Y | Xi) = f(Xi)
where f(Xi) denotes some function of the explanatory variable X. In our
example, E(Y | Xi) is a linear function of Xi. Equation (2.2.1) is known as the
conditional expectation function (CEF) or population regression
function (PRF) or population regression (PR) for short. It states merely
that the expected value of the distribution of Y given Xi is functionally
related to Xi.

E(Y | Xi) = β1 + β2Xi

where β1 and β2 are unknown but fixed parameters known as the


regression coefficients; β1 and β2 are also known as intercept and
slope coefficients, respectively.

THE MEANING It can be interpreted in two different ways.


OF THE TERM
LINEAR

Linearity in the The first and perhaps more “natural” meaning of linearity is that the
Variables conditional expectation of Y is a linear function of Xi. a regression function
such as E(Y | Xi) = β1 + β2Xi² is not a linear function because the variable
X appears with a power or index of 2

Linearity in the The second interpretation of linearity is that the conditional expectation of
Parameters Y, E(Y | Xi), is a linear function of the parameters, the β’s; it may or may not
be linear in the variable X. In this interpretation E(Y | Xi) = β1 + β2Xi² is a
linear (in the parameter) regression model.

Now consider the model E(Y | Xi) = β1 + β2²Xi . Now suppose X = 3; then
we obtain E(Y | Xi) = β1 + 3β2² , which is nonlinear in the parameter β2.
The preceding model is an example of a nonlinear (in the parameter)
regression model.
Rules for a
function to be a
Linear
Regression

STOCHASTIC Given the income level of Xi, an individual family’s consumption


SPECIFICATION expenditure is clustered around the average consumption of all families
OF PRF at that Xi, that is, around its conditional expectation. Therefore, we can
express the deviation of an individual Yi around its expected value as
follows:
ui = Yi − E(Y | Xi)
or
Yi = E(Y | Xi) + ui

where the deviation ui is an unobservable random variable taking positive


or negative values. Technically, ui is known as the stochastic disturbance
or stochastic error term.

ui Can be positive or negative


E(u) = 0

Yi = E(Y | Xi) + ui
Yi = β1 + β2Xi + ui

Now if we take the expected value on both sides, we obtain

E(Yi | Xi) = E[E(Y | Xi)] + E(ui | Xi)


= E(Y | Xi) + E(ui | Xi)
E(ui | Xi) = 0

THE Errors occur when we have either taken a smaller dataset or not accounted
SIGNIFICANCE all the important relevant variables.
OF THE
STOCHASTIC
DISTURBANCE
TERM

Vagueness of The theory, if any, determining the behavior of Y may be, and often is,
theory: incomplete.
Other variables could be incorporated.
Modified variables could be taken.

Unavailability of It is a common experience in empirical analysis that the data we would


data: ideally like to have often are not available
Increases the sample size of U.
Core variables vs Not including the core variables.
peripheral Relevant support variables for regression with causation.
variables:

Intrinsic If behavior is irrational.


randomness in
human behavior:

Poor proxy Using poor proxy variables when data of real variables is not available.
variables:

Principle of If our theory is not strong enough to suggest what other variables might be
parsimony: included, why introduce more variables? Let ui represent all other
variables. Of course, we should not exclude relevant and important
variables just to keep the regression model simple.

Wrong functional Sometimes ln is used in the function’s both sides without knowing the
form: data.

THE SAMPLE Regression function based on sample data


REGRESSION
FUNCTION (SRF)
Now, analogously to the PRF that underlies the population regression line,
we can develop the concept of the sample regression function (SRF) to
represent the sample regression line. The sample counterpart of may be
written as:

we can express the SRF (2.6.1) in its stochastic form as follows:


the PRF based on the SRF is at best an approximate one. This
approximation is shown diagrammatically in Figure 2.5. For X = Xi , we
have one (sample) observation Y = Yi . In terms of the SRF, the observed Yi
can be expressed as:

and in terms of the PRF, it can be expressed as:


EXERCISES

2.1 What is the conditional expectation function or the population regression


function?

Answer

2.2 What is the difference between the population and sample regression
functions? Is this a distinction without difference?

Answer

2.3 What is the role of the stochastic error term ui in regression analysis? What
is the difference between the stochastic error term and the residual, uˆi
(cap) ?

Answer

2.4 Why do we need regression analysis? Why not simply use the mean value
of the regressand as its best value?
Answer

2.5 What do we mean by a linear regression model?

Answer

2.6 Determine whether the following models are linear in the parameters, or
the variables, or both. Which of these models are linear regression
models?

Answer

2.7
Answer

2.8

Answer

2.9
Answer

2.10
Answer

2.11
Answer

2.12

Answer

2.13

Answer

2.14
Answer
2.15
Answer

2.16
Answer
CHAPTER # 3 TWO-VARIABLE REGRESSION MODEL: THE PROBLEM OF
ESTIMATION

Date : 26th, Mar, 2023

Estimation of two Our task is to estimate the population regression function (PRF) on the
variable basis of the sample regression function (SRF) as accurately as possible.
regression model we have discussed two generally used methods of estimation:
(1) ordinary least squares (OLS)
(2) maximum likelihood (ML).

Ordinary Least The method of ordinary least squares is attributed to Carl Friedrich Gauss,
Square Method: a German mathematician.
To understand this method, we first explain the leastsquares principle.
Recall the two-variable PRF:

The PRF is not directly observable. We estimate it from the SRF:

where Yˆ i is the estimated (conditional mean) value of Yi.

which shows that the uˆi (the residuals) are simply the differences between
the actual and estimated Y values.
We have to choose the SRF in such a way that the sum of the residuals
is as small as possible.
We adopt the least-squares criterion, which states that the SRF can be
fixed in such a way that

A further justification for the least-squares method lies in the fact that the
estimators obtained by it have some very desirable statistical properties,
as we shall see shortly. It is obvious that:

Where n is the sample size. These simultaneous equations are known as


the normal equations.
Solving the normal equations simultaneously, we obtain:
The estimators obtained previously are known as the least-squares
estimators, for they are derived from the least-squares principle.

THE
ASSUMPTIONS
UNDERLYING
THE METHOD OF
LEAST SQUARES

Assumption : 1 Linear regression model. The regression model is linear in the parameters.

Assumption : 2 X values are fixed in repeated sampling. Values taken by the regressor X
are considered fixed in repeated samples. More technically, X is assumed
to be nonstochastic.
Assumption : 3 Zero mean value of disturbance ui. Given the value of X, the mean, or
expected, value of the random disturbance term ui is zero. Technically, the
conditional mean value of ui is zero. Symbolically, we have:

Assumption : 4 Homoscedasticity or equal variance of ui. Given the value of X, the


variance of ui is the same for all observations. That is, the conditional
variances of ui are identical. Symbolically, we have:

where var stands for variance.

Assumption : 5 No autocorrelation between the disturbances. Given any two X values,


Xi and Xj (i ≠ j), the correlation between any two ui and uj (i ≠ j) is zero.
Symbolically,

where i and j are two different observations and where cov means
covariance.

Assumption : 6 Zero covariance between ui and Xi, or Formally,


Assumption : 7 The number of observations n must be greater than the number of
parameters to be estimated. Alternatively, the number of observations n
must be greater than the number of explanatory variables.

Assumption : 8 Variability in X values. The X values in a given sample must not all be the
same. Technically, var (X) must be a finite positive number.

Assumption : 9 The regression model is correctly specified. Alternatively, there is no


specification bias or error in the model used in empirical analysis.

Assumption : 10 There is no perfect multicollinearity. That is, there are no perfect linear
relationships among the explanatory variables.

PRECISION OR
STANDARD
ERRORS
OF
LEAST-SQUARES
ESTIMATES

Date: 8, May, 2023


PROPERTIES OF The least-squares estimates possess some ideal or optimum properties.
LEAST-SQUARES These properties are contained in the well-known Gauss–Markov
ESTIMATORS: theorem. To understand this theorem, we need to consider the best linear
THE unbiasedness property of an estimator.
GAUSS–MARKOV
An estimator, say the OLS estimator βˆ2, is said to be a best linear
THEOREM
unbiased estimator (BLUE) of β2 if the following hold:

1. It is linear, that is, a linear function of a random variable, such as the


dependent variable Y in the regression model.
2. It is unbiased, that is, its average or expected value, E(βˆ2), is equal to
the true value, β2.
3. It has minimum variance in the class of all such linear unbiased
estimators; an unbiased estimator with the least variance is known as an
efficient estimator
Date: 10, May, 2023

THE The coefficient of determination r² (two-variable case) or R² (multiple


COEFFICIENT OF regression) is a summary measure that tells how well the sample
DETERMINATION regression line fits the data. Also tells you variation in Y explained by X.
r²: A MEASURE
OF “GOODNESS
OF FIT”

(ESS) = Explained sum of squares


(RSS) = Residual sum of squares
(TSS) = Total sum of squares
r is a quantity closely related to but conceptually very much different from
r².

r can be +ve or -ve


r² lies between 0 and 1.
Properties of r Some of the properties of r are as follows:

Date: 11, May, 2023

Example:
HYPOTHETICAL
DATA ON WEEKLY
FAMILY
CONSUMPTION
EXPENDITURE Y
AND WEEKLY
FAMILY INCOME X
Monte Carlo The answer is provided by the so-called Monte Carlo experiments, which
Simulation: are essentially computer simulation, or sampling, experiments. To
introduce the basic ideas, consider our two-variable PRF:

EXERCISES
3.1

Answer

3.2
Answer

3.3

Answer
3.4

Answer

3.5

Answer
3.6

Answer

3.7

Answer

3.8

Answer
3.9
Answer

3.10
Answer

3.11

Answer
3.12

Answer

3.13
Answer

3.14

Answer
3.15

Answer

3.16
Answer

3.17

Answer
3.18
Answer

3.19

Answer
3.20
Answer
3.21
Answer

3.22
Answer

3.23
Answer

3.24

Answer

3.25
Answer

3.26

Answer
CHAPTER # 5 TWO-VARIABLE REGRESSION: INTERVAL ESTIMATION
AND HYPOTHESIS TESTING

Hypothesis ● To find out significance of βs


Testing ● To find out overall significance of the model

t distribution / t For finding the individual significance of the parameters.


test:

t-test
​ Null hypothesis (H₀): βˆ2 = 0
● This means the predictor variable has no significant effect on
the outcome variable.
​ Alternative hypothesis (H₁): βˆ2 ≠ 0
● This means the predictor variable has a significant effect on
the outcome variable.
​ T-statistic formula:
● t = (βˆ2 - 0) / SE(βˆ2) (where SE represents the standard error
of βˆ2)
​ Standard error (SE) formula:
● SE(βˆ2) = √[Σ(y - ŷ)² / (n - k - 1)] / √[Σ(x₂ - x̄₂)²] (where y
represents the observed outcome variable, ŷ represents the
predicted outcome variable, x₂ represents the values of the
predictor variable, x̄₂ represents the mean of the predictor
variable, n represents the sample size, and k represents the
number of predictors in the regression model)
​ Degrees of freedom (df):
● df = n - k - 1 (where n represents the sample size and k
represents the number of predictors in the regression model)
​ Interpretation:
● Calculate the t-statistic using the given formulas.
● Compare the t-statistic to the critical t-value corresponding to
the desired significance level and degrees of freedom.
● If the t-statistic is greater than the critical t-value (rejecting the
null hypothesis), it indicates that the predictor variable has a
significant effect on the outcome variable.
● If the t-statistic is not greater than the critical t-value (failing
to reject the null hypothesis), it suggests that the predictor
variable does not have a significant effect on the outcome
variable.

f-test
​ Null hypothesis (H₀): β₁ = β₂ = 0
● This means none of the predictor variables have a significant
effect on the outcome variable.
​ Alternative hypothesis (H₁): At least one of the β coefficients is not
equal to zero.
● This means at least one of the predictor variables has a
significant effect on the outcome variable.
​ F-statistic formula:
● F = [RSS / k] / [(TSS - RSS) / (n - k - 1)] (where RSS represents
the residual sum of squares, k represents the number of
predictors in the regression model, TSS represents the total
sum of squares, and n represents the sample size)
​ Residual sum of squares (RSS) formula:
● RSS = Σ(y - ŷ)² (where y represents the observed outcome
variable and ŷ represents the predicted outcome variable)
​ Total sum of squares (TSS) formula:
● TSS = Σ(y - ȳ)² (where y represents the observed outcome
variable and ȳ represents the mean of the outcome variable)
​ Degrees of freedom (df):
● df numerator = k
● df denominator = n - k - 1
​ Interpretation:
● Calculate the F-statistic using the given formulas.
● Compare the F-statistic to the critical F-value corresponding
to the desired significance level, degrees of freedom
numerator, and degrees of freedom denominator.
● If the F-statistic is greater than the critical F-value (rejecting
the null hypothesis), it indicates that at least one of the
predictor variables has a significant effect on the outcome
variable.
● If the F-statistic is not greater than the critical F-value (failing
to reject the null hypothesis), it suggests that none of the
predictor variables have a significant effect on the outcome
variable.

Overall
significance of ​ Review the t-test results:
the model ● Examine the t-test results for each parameter (β-coefficient) in
the regression model. Determine whether each predictor
variable is individually significant (i.e., reject or fail to reject
the null hypothesis for each parameter).
​ Consider the F-test result:
● Analyze the F-test result, which assesses the overall
significance of the regression model.
● If the F-statistic is greater than the critical F-value at the
desired significance level, it suggests that at least one of the
predictor variables has a significant effect on the outcome
variable, indicating overall significance of the regression
model.
​ Interpretation:
● If the F-test indicates overall significance (step 2) and at least
one of the t-tests for the individual parameters is significant
(step 1), you can conclude that the regression model is overall
significant, and one or more predictor variables have a
significant effect on the outcome variable.
● If the F-test is significant (overall significance) but all t-tests
for the individual parameters are not significant, it suggests
that the regression model is significant as a whole, but the
specific predictors may not be individually significant. In this
case, consider the practical implications and the context of
the analysis.
​ Consider other factors:
● Evaluate the adjusted R-squared value to assess the
proportion of variation in the outcome variable explained by
the regression model. A higher adjusted R-squared value
indicates a better fit and strengthens the overall significance.
Jarque–Bera (JB) The Jarque-Bera (JB) test is a statistical test used to assess the normality
Test of Normality:
of a data set based on its skewness and kurtosis. Here's an explanation of

the JB test in a similar format:

​ Null hypothesis (H₀): The data follows a normal distribution.


● This means the data is symmetrically distributed and has a
specific shape resembling a bell curve.
​ Alternative hypothesis (H₁): The data does not follow a normal
distribution.
● This means the data deviates from the assumptions of a
normal distribution, potentially exhibiting skewness or
kurtosis.
​ JB test statistic:
● JB = (n/6) * (S² + (K-3)²/4) (where n represents the sample
size, S represents the skewness coefficient, and K represents
the kurtosis coefficient)
​ Interpretation:
● Calculate the JB test statistic using the given formula.
● Compare the JB test statistic to the critical value
corresponding to the desired significance level.
● If the JB test statistic is greater than the critical value
(rejecting the null hypothesis), it indicates that the data
significantly deviates from a normal distribution.
● If the JB test statistic is not greater than the critical value
(failing to reject the null hypothesis), it suggests that the data
does not significantly deviate from a normal distribution.

Note: The critical value depends on the desired significance level and the

degrees of freedom, which is typically 2. In large sample sizes, the JB test

may be sensitive to small deviations from normality, so it is recommended

to consider other diagnostic tools and the practical context of the data

analysis.
The JB test helps in assessing the normality assumption, which is

important for many statistical analyses that rely on the assumption of

normal distribution.

Histogram of The Histogram of Residuals test is a visual method to assess the normality
Residuals assumption of the residuals in a statistical model. Here's an explanation of
the Histogram of Residuals test in a similar format:

​ Collect the residuals:


● Calculate the residuals by subtracting the predicted values
from the observed values in your regression model. Residuals
represent the differences between the observed data points
and the model's predictions.
​ Create a histogram:
● Construct a histogram using the residuals. A histogram is a
graphical representation of the distribution of data, where the
data range is divided into bins or intervals, and the height of
each bin represents the frequency of data falling within that
range.
​ Assess the shape:
● Analyze the shape of the histogram of residuals. Look for
characteristics that indicate departure from a normal
distribution, such as skewness (asymmetry) or heavy tails.
​ Interpretation:
● If the histogram of residuals appears approximately
bell-shaped, symmetric, and centered around zero, it suggests
that the residuals follow a normal distribution, supporting the
normality assumption.
● If the histogram of residuals shows substantial skewness
(shift to one side) or has heavy tails (data points extending far
from the center), it indicates a departure from normality. This
suggests that the normality assumption may not hold for the
residuals.

Note: The Histogram of Residuals test is a visual method and does not
provide formal statistical measures like p-values or test statistics. It is
mainly used as a diagnostic tool to identify potential departures from
normality in the residuals. If a significant departure from normality is
observed, it may be necessary to explore alternative statistical techniques
or consider data transformations to address the violation of the normality
assumption.
Normal The Normal Probability Plot, also known as the Q-Q plot (quantile-quantile
Probability Plot plot), is a graphical method to assess the normality assumption of a data
set. Here's an explanation of the Normal Probability Plot test in a similar
format:

​ Collect the residuals:


● Calculate the residuals by subtracting the predicted values
from the observed values in your regression model. Residuals
represent the differences between the observed data points
and the model's predictions.
​ Plot the Normal Probability Plot:
● Plot the residuals on the y-axis and their expected quantiles
(based on a normal distribution) on the x-axis. Each data point
represents the quantiles of the residuals.
​ Assess the linearity:
● Examine the pattern formed by the plotted points. If the points
closely follow a straight line, it indicates that the residuals are
normally distributed. Deviations from a straight line suggest
departures from normality.
​ Interpretation:
● If the points on the Normal Probability Plot align closely along
a straight line, it suggests that the residuals follow a normal
distribution, supporting the normality assumption.
● If the points on the plot deviate from a straight line, exhibiting
curvature, bends, or deviations at the tails, it indicates a
departure from normality. This suggests that the normality
assumption may not hold for the residuals.

Note: The Normal Probability Plot is a visual method and does not provide
formal statistical measures like p-values or test statistics. It serves as a
diagnostic tool to assess departures from normality. If a significant
departure from normality is observed, it may be necessary to explore
alternative statistical techniques or consider data transformations to
address the violation of the normality assumption.

EXERCISES
5.1

Answer
5.2
Answer

5.3
Answer

5.4

Answer
5.5

Answer
5.6

Answer

5.7
Answer
5.8
Answer
5.9
Answer
5.10

Answer
5.11
Answer
5.12

Answer
5.13
Answer
5.14

Answer

5.15
Answer

5.16
Answer
5.17
Answer

5.18

Answer
5.19
Answer
CHAPTER # 6 EXTENSIONS OF THE TWO-VARIABLE LINEAR
REGRESSION MODEL

31, May, 2023

REGRESSION There are occasions when the two-variable PRF assumes the following
THROUGH THE form:
ORIGIN

In this model the intercept term is absent or zero, hence the name
regression through the origin.
As an illustration, consider the Capital Asset Pricing Model (CAPM) of
modern portfolio theory, which, in its risk-premium form, may be
expressed as:
FUNCTIONAL In particular, we discuss the following regression models:
FORMS OF 1. The log-linear model
REGRESSION 2. Semilog models
MODELS 3. Reciprocal models
4. The logarithmic reciprocal model
Exponential
regression model
SEMILOG You may recall the following well-known compound interest formula from
MODELS: your introductory course in economics.
LOG–LIN AND
LIN–LOG
MODELS where r is the compound (i.e., over time) rate of growth of Y. Taking the
natural logarithm of, we can write

Linear Trend Instead of estimating model researchers sometimes estimate the following
Model model:
That is, instead of regressing the log of Y on time, they regress Y on time,
where Y is the regressand under consideration. Such a model is called a
linear trend model and the time variable t is known as the trend variable. If
the slope coefficient in is positive, there is an upward trend in Y, whereas if
it is negative, there is a downward trend in Y. For the expenditure on
services data that we considered earlier, the results of fitting the linear
trend model are as follows:

The Lin–Log Unlike the growth model just discussed, in which we were interested in
Model finding the percent growth in Y for an absolute change in X, suppose we
now want to find the absolute change in Y for a percent change in X. A
model that can accomplish this purpose can be written as:

For descriptive purposes we call such a model a lin–log model. Let us


interpret the slope coefficient β2. As usual,

The second step follows from the fact that a change in the log of a number
is a relative change.

Symbolically, we have
where, as usual, delta denotes a small change. Equationcan be written,
equivalently, as

RECIPROCAL Models of the following type are known as reciprocal models.


MODELS
EXERCISES

6.1

Answer

6.2
Answer
6.3

Answer

6.4
Answer

6.5
Answer

6.6
Answer

6.7

Answer

6.8
Answer

6.9

Answer

6.10

Answer
6.11

Answer

6.12
Answer

6.13

Answer
6.14

Answer

6.15
Answer

6.16
Answer

6.17

Answer
6.18

Answer

6.19
Answer
CHAPTER # 7 MULTIPLE REGRESSION ANALYSIS: THE PROBLEM OF
ESTIMATION

14, June, 2023

The three Generalizing the two-variable population regression function (PRF)


variable model

INTERPRETATIO Given the assumptions of the classical regression model, it follows that, on
N OF MULTIPLE taking the conditional expectation of Y on both sides
REGRESSION
EQUATION
In words, this equation gives the conditional mean or expected value of Y
conditional upon the given or fixed values of X2 and X3. Therefore, as in the
two-variable case, multiple regression analysis is regression analysis
conditional upon the fixed values of the regressors, and what we obtain is
the average or mean value of Y or the mean response of Y for the given
values of the regressors.

OLS Estimators To find the OLS estimators, let us first write the sample regression function
(SRF) corresponding to the PRF
As noted in Chapter 3, the OLS procedure consists in so choosing the
values of the unknown parameters that the residual sum of squares (RSS)
is as small as possible. Symbolically,

Variances and
Standard Errors
of OLS
Estimators

where r23 is the sample coefficient of correlation between X2 and X3 as


defined in Chapter 3.
In all these formulas σ2 is the (homoscedastic) variance of the population
disturbances ui.

19, June, 2023

THE MULTIPLE By definition.


COEFFICIENT OF
DETERMINATION

Range: 0 - 1

R² AND THE
ADJUSTED R²
THE The Cobb-Douglas production function is a mathematical representation of
COBB–DOUGLAS the relationship between output and inputs in production theory. In its
PRODUCTION stochastic form, it can be expressed as:
FUNCTION

To convert this nonlinear relationship into a linear one, we can


log-transform the model:

This transformation makes the model linear in the parameters β0, β2, and
β3. However, it remains nonlinear in the variables Y and X, but linear in
their logarithms. This log-log model is the multiple regression counterpart
of the two-variable log-linear model.

The properties of the Cobb-Douglas production function are as follows:

β2 represents the elasticity of output with respect to the labor input, while
β3 represents the elasticity of output with respect to the capital input.
The sum of β2 and β3 provides information about the returns to scale. If
the sum is 1, there are constant returns to scale. If the sum is less than 1,
there are decreasing returns to scale, and if the sum is greater than 1, there
are increasing returns to scale.
An example of the Cobb-Douglas production function is provided using
data from the agricultural sector of Taiwan for the period 1958-1972. By
applying the Ordinary Least Squares (OLS) method, the regression results
are as follows:

ln Yi = -3.3384 + 1.4988 ln X2i + 0.4899 ln X3i

The coefficients 1.4988 and 0.4899 represent the output elasticities of


labor and capital, respectively. They indicate the percentage change in
output for a 1 percent change in the corresponding input, holding the other
input constant. The sum of these coefficients, 1.9887, represents the
returns to scale parameter, indicating increasing returns to scale. The
R-squared value of 0.8890 indicates that approximately 89 percent of the
variation in the log of output can be explained by the logs of labor and
capital inputs.

These regression results can be further analyzed using statistical methods


to test hypotheses about the parameters of the Cobb-Douglas production
function for the Taiwanese economy.

PARTIAL Partial correlation coefficients measure the strength and direction of the
CORRELATION linear relationship between two variables while controlling for the influence
COEFFICIENTS of one or more additional variables. In other words, they quantify the
correlation between two variables after removing the effects of other
variables.

These partial correlations can be easily obtained from the simple or zero
order, correlation coefficients as follows
The partial correlations given in Eqs. are called first order correlation
coefficients.
Thus r12.34 would be the correlation coefficient of order two, r12.345 would be
the correlation coefficient of order three, and so on.

EXERCISES

7.1
Answer

7.2

Answer
7.3

Answer

7.4

Answer
7.5

Answer

7.6

Answer

7.7
Answer

7.8

Answer

7.9

Answer
7.10

Answer

7.11

Answer

7.12
Answer

7.13

Answer
7.14

Answer
7.15

Answer
7.16

Answer
7.17

Answer
7.18

Answer
7.19
Answer
7.20
Answer

7.21
Answer
7.22

Answer
7.23
Answer

7.24
Answer
CHAPTER # 8 MULTIPLE REGRESSION ANALYSIS: THE PROBLEM OF
INFERENCE

21, June, 2023

Testing the
Overall
Significance of a
Multiple
Regression in
Terms of R²

R² and F have a positive relationship,

Uses of F test
Overall For checking the overall significance.
Significance To test the overall significance of a sample regression using the F-test, we
assess whether the regression model as a whole is statistically significant
in explaining the dependent variable. Here's how it can be done:

​ Start with the sample regression model: Y = β0 + β1X1 + β2X2 + ... +


βnXn + ε, where Y is the dependent variable, X1, X2, ..., Xn are the
independent variables, β0, β1, β2, ..., βn are the regression
coefficients, and ε is the error term.
​ Calculate the sum of squares regression (SSR), which represents the
variation explained by the regression model, using the formula: SSR
= ∑(Ŷi - Ȳ)^2, where Ŷi is the predicted value of Y for each
observation i, and Ȳ is the mean of the dependent variable Y.
​ Calculate the degrees of freedom for the regression (dfR), which is
equal to the number of independent variables in the model.
​ Calculate the sum of squares error (SSE), which represents the
unexplained variation or residual variation, using the formula: SSE =
∑(Yi - Ŷi)^2, where Yi is the actual value of Y for each observation i.
​ Calculate the degrees of freedom for the error (dfE), which is equal
to the total number of observations minus the number of
independent variables minus 1.
​ Calculate the F statistic using the formula: F = (SSR / dfR) / (SSE /
dfE). The F statistic measures the ratio of the explained variation to
the unexplained variation, adjusted for the degrees of freedom.
​ Determine the critical value of the F statistic based on the desired
significance level (e.g., α = 0.05) and the degrees of freedom (dfR
and dfE).
​ Compare the calculated F statistic to the critical value. If the
calculated F statistic is greater than the critical value, we reject the
null hypothesis, indicating that the regression model as a whole is
statistically significant.
​ Alternatively, we can calculate the p-value associated with the F
statistic using the F-distribution. If the p-value is less than the
chosen significance level (e.g., p < 0.05), we reject the null
hypothesis and conclude that the regression model is statistically
significant.

In summary, the F-test for the overall significance of a sample regression


involves calculating the F statistic by comparing the explained and
unexplained variations in the model. By comparing the F statistic to the
critical value or calculating the p-value, we can determine if the regression
model is statistically significant in explaining the dependent variable.
Testing Linear The F-test is commonly used to test the linear equality restriction in a
equality regression model, which means comparing two or more linear regression
restrictions: models to determine if they are statistically different. Here's how it can be
done:

​ Start with the unrestricted regression model, which includes all the
variables and parameters of interest: Y = β0 + β1X1 + β2X2 + ... +
βnXn + ε, where Y is the dependent variable, X1, X2, ..., Xn are the
independent variables, β0, β1, β2, ..., βn are the regression
coefficients, and ε is the error term.
​ Create the restricted regression model by imposing the linear
equality restriction. For example, if we want to test if the coefficients
of X1 and X2 are equal, the restricted model would be: Y = β0 + β1X1
+ β1X2 + ... + βnXn + ε.
​ Calculate the sum of squares error (SSE) for both the unrestricted
and restricted models.
​ Calculate the degrees of freedom for the error for both models. The
degrees of freedom are determined by the total number of
observations minus the number of parameters estimated in the
model.
​ Use the SSE and degrees of freedom to calculate the F statistic
using the formula: F = ((SSE_restricted - SSE_unrestricted) /
(df_restricted - df_unrestricted)) / (SSE_unrestricted /
df_unrestricted).
​ Determine the critical value of the F statistic based on the desired
significance level (e.g., α = 0.05) and the degrees of freedom.
​ Compare the calculated F statistic to the critical value. If the
calculated F statistic is greater than the critical value, we reject the
null hypothesis, indicating that the restricted model is statistically
different from the unrestricted model.
​ Alternatively, we can calculate the p-value associated with the F
statistic using the F-distribution. If the p-value is less than the
chosen significance level (e.g., p < 0.05), we reject the null
hypothesis and conclude that the restricted model is statistically
different from the unrestricted model.

In summary, the F-test for testing linear equality restrictions involves


comparing the restricted and unrestricted regression models using the F
statistic. By comparing the F statistic to the critical value or calculating the
p-value, we can determine if the linear equality restriction holds or if the
restricted model is statistically different from the unrestricted model.
Check Structured The F-test can also be used to check for a structured break in a regression
break: model, which involves examining whether there is a significant change in
the relationship between the dependent variable and independent variables
at a specific point in the data. Here's how it can be done:

​ Divide the dataset into two groups: the pre-break period and the
post-break period. The break point should be selected based on
theoretical considerations or prior knowledge.
​ Fit two separate regression models: one for the pre-break period and
one for the post-break period. Let's denote these models as Model 1
and Model 2, respectively.
​ Calculate the sum of squares error (SSE) for each model.
​ Calculate the degrees of freedom for the error for each model. The
degrees of freedom are determined by the total number of
observations minus the number of parameters estimated in the
model.
​ Use the SSE and degrees of freedom to calculate the F statistic
using the formula: F = ((SSE1 - SSE2) / (df1 - df2)) / (SSE2 / df2).
​ Determine the critical value of the F statistic based on the desired
significance level (e.g., α = 0.05) and the degrees of freedom.
​ Compare the calculated F statistic to the critical value. If the
calculated F statistic is greater than the critical value, we reject the
null hypothesis, indicating a significant structured break in the
relationship between the variables.
​ Alternatively, we can calculate the p-value associated with the F
statistic using the F-distribution. If the p-value is less than the
chosen significance level (e.g., p < 0.05), we reject the null
hypothesis and conclude that there is a significant structured break.

In summary, the F-test for checking a structured break involves fitting


separate regression models for the pre-break and post-break periods and
comparing their SSE values using the F statistic. By comparing the F
statistic to the critical value or calculating the p-value, we can determine if
there is a significant change in the relationship between the variables at the
break point.
CHAPTER # 10 MULTICOLLINEARITY: WHAT HAPPENS IF THE
REGRESSORS ARE CORRELATED?

26, June, 2023

Multicollinearity Multicollinearity refers to a situation in which two or more independent


variables in a regression or statistical model are highly correlated with each
other. It indicates a high degree of linear relationship between the
predictors.

Sources of Linear Relationships: Multicollinearity can arise when there are strong
Multicollinearity linear relationships among the independent variables. For example, if two
variables are measuring similar aspects of the same phenomenon or if one
variable can be expressed as a linear combination of others, it can lead to
high correlation and multicollinearity.

Measurement Errors: Inaccurate or imprecise measurement of variables


can introduce noise and random errors, which may lead to spuriously high
correlations between the variables. Measurement errors can contribute to
multicollinearity, especially when multiple variables are measured with
similar types of errors.

Data Transformation: Applying mathematical transformations to variables,


such as taking logarithms or squaring, can sometimes induce
multicollinearity. Certain transformations can amplify the relationships
between variables and increase correlation, leading to multicollinearity.

Aggregation or Collapsing: Combining or aggregating variables that are


already correlated can introduce multicollinearity. For example, averaging
or summing variables that measure similar concepts can create high
correlation and multicollinearity.

Overlapping Data Sources: When using multiple data sources that


contain similar or redundant information, it can lead to multicollinearity. If
the different data sources share common underlying factors or are based
on overlapping measurements, it can result in high correlation and
multicollinearity.

Small Sample Size: In smaller datasets, chance fluctuations or random


sampling variation can lead to spurious correlation between variables,
contributing to multicollinearity.
For K variables Multicollinearity occurs when there is a high correlation among the
independent variables in a model with K variables, making it difficult to
separate their individual effects.

Inverse of VIF is VIF measures the extent to which the variance of the estimated regression
tolerance (Tol) coefficient is inflated due to multicollinearity. A VIF value greater than 1
suggests the presence of multicollinearity. Generally, VIF values above 5 or
10 are considered problematic. For instance, a VIF of 8 for Variable 1
indicates that the variance of its estimated coefficient is significantly
inflated due to multicollinearity.

Auxillary Since multicollinearity arises because one or


Regression more of the regressors are exact or approximately linear combinations of
the other regressors, one way of finding out which X variable is related to
other X variables is to regress each Xi on the remaining X variables and
compute the corresponding R2, which we designate as R2
i ; each one of these regressions is called an auxiliary regression, auxiliary
to the main regression
of Y on the X’s. Then, following the relationship between F and R2
established the variable:

Klien’s Instead of formally testing all auxiliary R2 values, one may adopt Klien’s
rule of thumb rule of thumb, which suggests that multicollinearity may be a troublesome
problem only if the R2 obtained from an auxiliary regression is greater than
the overall R2, that is, that obtained from the regression of Y on all the
regressors. Of course, like all other rules of thumb, this one should be used
judiciously.
CHAPTER # 12 AUTOCORRELATION: WHAT HAPPENS IF THE ERROR
TERMS ARE CORRELATED?

Autocorrelation Autocorrelation, also known as serial correlation, refers to the correlation


between a variable and its lagged values. It indicates the degree to which
the current value of a variable is related to its past values. Autocorrelation
can be positive (values are similar across time) or negative (values alternate
across time).

Sources of Autocorrelation can arise due to various factors, such as omitted variables,
Autocorrelation: misspecified models, time-dependent trends, or the presence of
seasonality. For example, if a time series variable exhibits a consistent
increasing or decreasing trend over time, it may lead to positive
autocorrelation.

Specification Specification bias refers to the incorrect specification of a regression


Biased: model, which can lead to biased and inconsistent estimates of the
coefficients. Autocorrelation can introduce specification bias when it is not
adequately accounted for in the model. Failing to include lagged values of
the dependent variable or relevant independent variables can result in
specification bias.

Lags: Lags refer to the number of previous time periods considered when
calculating autocorrelation. It represents the time lag between the current
observation and the past observation being compared. For example, a lag
of 1 represents the correlation between a variable's current value and its
immediately preceding value.

Detection Autocorrelation can be detected using various statistical methods, such as


method of the Durbin-Watson test, Ljung-Box test, or examining autocorrelation plots.
Autocorrelation: These methods assess the degree of correlation between residuals or the
dependent variable and its lagged values. The Durbin-Watson test, for
example, tests for the presence of autocorrelation by evaluating the serial
correlation of residuals in a regression model.

Run test The Run Test is a nonparametric test used to examine the presence of runs
or patterns in a sequence of binary or categorical data. It analyzes whether
the data is randomly distributed or shows systematic patterns. The test is
based on the number of runs (sequences of consecutive similar values)
observed in the data compared to the number of runs expected under the
assumption of randomness.

Numerically, the Run Test involves the following steps:

Define the null hypothesis (H0): The null hypothesis assumes randomness
in the data, indicating that there is no systematic pattern or trend.

Calculate the observed number of runs (Robs): Count the number of runs
(sequences of consecutive similar values) in the data sequence.

Calculate the expected number of runs (Rexp): The expected number of


runs is determined based on the formula:

Rexp = 1 + (2 * N1 * N2) / (N1 + N2)

where N1 is the total number of positive runs (sequences of 1s) and N2 is


the total number of negative runs (sequences of 0s).

Calculate the test statistic (Z-score): The test statistic is calculated using
the formula:

Z = (Robs - Rexp) / sqrt(V)

where V = (Robs + Rexp + 1) * (Robs + Rexp) / (N1 + N2 + 1)

The Z-score measures the deviation of the observed number of runs from
the expected number of runs, taking into account the variance of the runs.

Compare the Z-score to the critical values: The Z-score is compared to


critical values (typically based on a standard normal distribution) to
determine the statistical significance. If the absolute value of the Z-score
exceeds the critical value, the null hypothesis is rejected, suggesting the
presence of non-randomness or a systematic pattern in the data.

Durbin Watson The Durbin-Watson test is a statistical test used to assess the presence of
d-test autocorrelation in the residuals of a regression model. Autocorrelation in
the residuals indicates a systematic pattern or correlation between the
error terms, which violates the assumption of independence.
Numerically, the Durbin-Watson test involves the following steps:

Define the null hypothesis (H0): The null hypothesis assumes no


autocorrelation in the residuals, indicating that the error terms are
independent.

Calculate the Durbin-Watson statistic (d): The Durbin-Watson statistic is


calculated using the formula:

d = (Σ(ei - ei-1)²) / Σei²

where ei represents the residuals at time i and ei-1 represents the residuals
at the previous time point.

Note: The calculation involves summing the squared differences between


consecutive residuals and dividing it by the sum of squared residuals.

Determine the critical values: The critical values of the Durbin-Watson


statistic depend on the significance level, the number of observations, and
the degrees of freedom. Tables or statistical software provide critical
values for different significance levels and sample sizes.

Compare the Durbin-Watson statistic to the critical values: The


Durbin-Watson statistic ranges from 0 to 4. A value close to 2 indicates no
significant autocorrelation, while values below 2 suggest positive
autocorrelation, and values above 2 suggest negative autocorrelation. To
determine significance, the Durbin-Watson statistic is compared to the
critical values. If the statistic falls outside the critical value range, the null
hypothesis is rejected, indicating the presence of autocorrelation.

By calculating the Durbin-Watson statistic based on the formula and


comparing it to the critical values, the Durbin-Watson test allows you to
assess whether autocorrelation exists in the residuals of a regression
model.

Assumptions of d The Durbin-Watson test relies on several assumptions to provide accurate


- test and meaningful results. These assumptions are crucial for interpreting the
test statistic correctly. Here are the key assumptions of the Durbin-Watson
test:
Linearity: The Durbin-Watson test assumes a linear relationship between
the dependent variable and the independent variables. The test is primarily
designed for linear regression models. If the relationship is nonlinear, the
test results may not be valid or informative.

Independence: The test assumes that the residuals (or error terms) in the
regression model are independent of each other. Autocorrelation in the
residuals violates this assumption and can impact the validity of the test.
The presence of autocorrelation undermines the independence
assumption and may lead to biased results.

No perfect multicollinearity: The Durbin-Watson test assumes that there is


no perfect multicollinearity among the independent variables in the
regression model. Perfect multicollinearity occurs when one or more
independent variables can be perfectly predicted by a linear combination
of the other variables. It makes the estimation of regression coefficients
impossible and affects the interpretation of the Durbin-Watson test.

Homoscedasticity: Homoscedasticity assumes that the variance of the


residuals is constant across all levels of the independent variables. If there
is heteroscedasticity present in the data, where the variance of the
residuals varies systematically, the Durbin-Watson test results may not be
reliable.

Absence of outliers: The Durbin-Watson test assumes that there are no


influential outliers in the data that would significantly affect the results.
Outliers can distort the residuals and potentially invalidate the test
outcomes.

It is important to check these assumptions before applying the


Durbin-Watson test. Violations of these assumptions may require
alternative diagnostic tests or the use of alternative regression models that
account for the specific issues present in the data.

The Durbin h The Durbin h-statistics, also known as Durbin's alternative test statistics,
statistics are alternative measures used to test for autocorrelation in the residuals of
a regression model. These statistics provide an alternative approach to
detect autocorrelation when the assumptions of the Durbin-Watson test
are not met.
Numerically, the Durbin h-statistics involve the following steps:

Calculate the residuals: Calculate the residuals (e) from the regression
model, which represent the differences between the observed values and
the predicted values.

Calculate the test statistics (h-statistics): There are multiple h-statistics,


denoted as h(0), h(1), h(2), etc., each corresponding to a different lag order.
The h-statistics are calculated using the formula:

h(k) = Σ(ei - ei-k)² / Σei²

where k represents the lag order, ei represents the residuals at time i, and
ei-k represents the residuals at a lagged time point.

Note: The calculation involves summing the squared differences between


residuals at different lags and dividing it by the sum of squared residuals.

Determine the critical values: The critical values for the h-statistics depend
on the significance level, the sample size, and the number of parameters in
the regression model. Critical values can be obtained from statistical tables
or software.

Compare the h-statistics to the critical values: The h-statistics are


compared to the critical values to determine statistical significance. If the
h-statistic exceeds the critical value, it suggests the presence of
autocorrelation. The specific lag order with significant autocorrelation can
be identified by examining which h-statistics exceed their respective
critical values.

By calculating the h-statistics based on the formula and comparing them


to the critical values, the Durbin h-statistics test allows you to assess
autocorrelation in the residuals of a regression model, providing an
alternative approach when the assumptions of the Durbin-Watson test are
not met.

BG-LM The Breusch-Godfrey LM (BG-LM) test, also known as the LM test for
autocorrelation, is a statistical test used to detect the presence of
autocorrelation in the residuals of a regression model. It is an extension of
the Durbin-Watson test and can be applied to both linear and nonlinear
regression models.

Numerically, the BG-LM test involves the following steps:

Specify the null hypothesis (H0): The null hypothesis assumes no


autocorrelation in the residuals, indicating that the error terms are
independent.

Estimate the regression model: Estimate the regression model and obtain
the residuals.

Extend the model with lagged residuals: Add lagged residuals as additional
explanatory variables to the model. The number of lagged residuals to
include is determined based on the desired lag order (e.g., 1, 2, 3).

Estimate the extended model: Estimate the extended model with lagged
residuals included and obtain the residuals.

Calculate the test statistic (LM statistic): The LM statistic is calculated by


regressing the residuals from the extended model on the lagged residuals
and obtaining the R-squared value. The LM statistic is equal to the sample
size multiplied by the R-squared value.

Determine the critical value: The critical value for the LM statistic depends
on the significance level and the degrees of freedom, which is equal to the
number of lagged residuals included in the extended model.

Compare the LM statistic to the critical value: If the LM statistic exceeds


the critical value, the null hypothesis is rejected, suggesting the presence
of autocorrelation in the residuals.

By estimating the regression model, extending it with lagged residuals,


calculating the LM statistic, and comparing it to the critical value, the
BG-LM test allows you to assess whether there is autocorrelation in the
residuals of the regression model.

Removal of AC The Removal of Autocorrelation (AC) test, also known as the


test Cochrane-Orcutt procedure, is a technique used to address the issue of
autocorrelation in a regression model by transforming the data to eliminate
the autocorrelation in the residuals. It is commonly applied when the
Durbin-Watson test or other diagnostics indicate the presence of
autocorrelation.

Numerically, the Removal of AC test involves the following steps:

Estimate the initial regression model: Begin by estimating the initial


regression model using the ordinary least squares (OLS) method.

Calculate the residuals: Calculate the residuals of the initial model by


subtracting the predicted values from the observed values.

Estimate the auxiliary regression model: Regress the residuals on the


lagged residuals. Include the lagged residuals as additional independent
variables in the model.

Calculate the coefficient estimate: Obtain the coefficient estimate for the
lagged residuals from the auxiliary regression model.

Calculate the correction factor: Calculate the correction factor as 1 minus


the coefficient estimate obtained in step 4.

Correct the initial regression model: Multiply the residuals from the initial
model by the correction factor calculated in step 5.

Re-estimate the corrected regression model: Estimate a new regression


model using the corrected residuals from step 6.

Assess the autocorrelation: Evaluate the residuals of the corrected model


to determine if the autocorrelation has been adequately addressed. This
can be done by conducting the Durbin-Watson test or examining other
diagnostic tests.

By iteratively applying the Removal of AC test, you can attempt to


eliminate autocorrelation in the residuals of the regression model. The
procedure involves estimating the initial model, calculating residuals,
estimating an auxiliary model, obtaining a correction factor, correcting the
initial model, and re-estimating the corrected model. This iterative process
is repeated until the autocorrelation in the residuals is sufficiently reduced
or eliminated.

You might also like