Scribd
Upload
7 views7 pages

Theme 4 Non Linear Models

The document discusses nonlinear regression models in econometrics, focusing on quadratic equations, dummy variables, interaction terms, and qualitative variables. It explains the use of logarithmic transformations to analyze data and the importance of interaction terms in modeling non-linear relationships. Additionally, it highlights the challenges of incorporating qualitative information and ordinal variables into econometric models.

Uploaded by

mismail10001000
Copyright
© © All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as DOCX, PDF, TXT or read online on Scribd
7 views7 pages

Theme 4 Non Linear Models

The document discusses nonlinear regression models in econometrics, focusing on quadratic equations, dummy variables, interaction terms, and qualitative variables. It explains the use of logarithmic transformations to analyze data and the importance of interaction terms in modeling non-linear relationships. Additionally, it highlights the challenges of incorporating qualitative information and ordinal variables into econometric models.

Uploaded by

mismail10001000
Copyright
© © All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as DOCX, PDF, TXT or read online on Scribd
You are on page 1/ 7

Theme 4: Nonlinear regression models – quadratic equations- dummy variable-

interaction term – qualitative variables and applications

4.1Nonlinear Models in Econometrics :

Ø It involves data scaling, as it involves the rescaling of the coefficients , standard


errors, confidence interval t statistics and F statistics. The question then that poses
itself is the following: -
Ø Example: A nonlinear variable would be dollars wage per additional year of
education. Labor and capital on productivity.

Ø Example: The effect of accumulated interest as of the book value of an asset on how
to calculate depreciation.
Ø If we were to measure, re-scale or take logs of the x and y in different units, only the
magnitudes of these derivatives would change but overall fit of the regression
equation wouldn’t. Thus, each coefficient’s t− statistic will have the same value, with
the same p− values, irrespective of scaling and R2as well will remain the same.

β 1−β
Q=a . L K

4.2.1 Using Logarithmic Functional Forms


Ø Even though the dependent variable might appear in a logarthmic form, yet changing
the untis and linearlizing it doesn’t appear to affect the coefficient’s slope.
As per log properities :-
Log (C 1 y i)= Log (C 1)+Log ( y i) for any constant C 1>1

Ø The standardization of variables “z transformed.” This transformation (subtracting the


mean and dividing by the sample standard deviation) generates variables with a mean
of zero and a standard deviation of one.
It is often used in this sort of model and analysis to estimate an exponential trend like growth
rate –since if the X variable is t, we have ∂log(y)/∂t, or an estimate of the growth rate of y.

Another example of log functions is the Cobb Douglas Function where


Y= A. K α L1−α= log (y) = log (A)+ α Log (K) + (1- α ¿ log (L)

Ø We recognize that equation is an exponential function of the form, y = a b x


Ø where in this case y =D (depreciation)
Ø a = B (book value)
Ø b= 1-i (1-i) and finally n = x (number of exponents quadratic, cubic. Exponential
progression….)
Ø On applying semi logarithmic transformation on the equation as follows: -
Ø The resulting linear form is as follows:
Log D = log B + n log (1-i)

Ø What are the benefits of the log ? What are their uses ?
1) Taking the log of a model doesn’t affect its variables since it measures percentage
changes
2) The intepretation of elasticity value is simple
3) Reduces heteroskedasticity, variability of data with high variances and skewness

Ø What is the difference between taking logs and levels in variables ?


Log Levels
Variables Ex: Population , GDP, trade Ex:, percentages growth rates
flows, wage ,firm’s sales and of GDP, inflation rate ,
market value. education , experience, tenure
Measurement Ex: big values that are A variable with measurement
measured in monetary units between [0-1] such as indices
such as dollars or in years / measured in
porportion or percent

4.3 Models with Quadratics

y= β❑0 + β❑ ❑ 2
1 x + β 2 x +u

to rewrite it as an estimate y Λ = β 0Λ + β 1Λ x 1Λ + β 2Λ x 2Λ 2
Linearlized formate and solving for effect of x 2Λ 2 on y Λ
Λ
∆ y ≈¿ )
Ø If y is regressed on x 1and x2, it is important to note that ∂y/∂x must be calculated
first by taking account of this form. (i.e.: we cannot consider the effect of changing x
while holding x2 constant).

4.4The Nature of Qualitative Information


Sometimes we can not obtain a set of numerical values for all the variables we want to
use in a model.

This is because some variables can not be quantified easily.

Examples:

Gender may play a role in determining salary levels

Different ethnic groups may follow different consumption patterns

Educational levels can affect earnings from employment


4.5 Interaction terms

Ø An important technique that allows for non- linearities in an econometric model is the
use of interaction terms–the product of explanatory variables. When a model with two
explanatory variables and their interaction terms is considered
Y= β 0+ β 1 x 1+ β 2 x 2+ β 3 x 1 x 2+u
Ø It is important to note that the presence of an interaction term changes the ’usual’
interpretation of the coefficients associated with the components of the interaction.

4.5.1 Using Dummy Variables for Multiple Categories

Ø Dummy variables are added to the independent variables and any categorical variable
can be turned into a set of dummy variables or dummies for sectors (nominal ones)
Ø For dummy variables, the base group is represented by the intercept, if there are n
categories there should be n – 1 dummy variable to avoid the dummy trap, which is
the situation of perfect collinearity and e-views will not complete the estimation.

Ø The model below consists of y (GPA), independent variables and dummies are:- high
schoool graduates (hsgrad.) , colleage graduates (colgrad.),male, female and crossing
and creating interaction terms between the variables to estimate difference in salary.

Formally, the model is y = b + d male + d hsgrad + d colgrad + d male*hsgrad + d male*colgrad + b x + ui


0 1 2 3 4 5 1

Ø The model below consists of y (GPA), independent variables and dummies are:- high schoool
graduates (hsgrad.) , colleage graduates (colgrad.),male, female and crossing and creating
interaction terms between the variables to estimate difference in salary

Formally, the model is y = b0 + d1male + d2hsgrad + d3colgrad + d4male*hsgrad +


d5male*colgrad + b1x + ui, then, for example:
If male = 0 and hsgrad = 0 and colgrad = 0
y = b0 + b1x + ui
If male = 0 and hsgrad = 1 and colgrad = 0
y = b0 + d2hsgrad + b1x + ui
If male = 1 and hsgrad = 0 and colgrad = 1
y = b0 + d1male + d3colgrad + d5male*colgrad + b1x +ui

4.5.2 Incorporating Ordinal Information through Dummies

Ø Ordinal varaible is a variable taking several rankings [0- 5] and each rank
differ in meaning than the other one. Ranked from worst being 0 and best 4.
Ø Ex: Moody’s investors services and Standard and Poor’s want to rank the
quality of debts for local governments based on investors’ behavior and
preferences like probability of default, interest rates.
Ø The ordinal variables are more often used in questionnaires and are called
Likert scale variables.

You might also like