Scribd
Upload
7 views17 pages

Session 3

The document discusses the Simple Linear Regression Model, focusing on the relationship between the dependent variable (y) and the independent variable (x). It outlines key issues such as allowing for other factors affecting y, understanding the functional relationship, and ensuring a ceteris paribus condition. Additionally, it covers assumptions related to the error term (u) and the derivation of Ordinary Least Squares (OLS) estimates.

Uploaded by

Pavneet Singh
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
7 views17 pages

Session 3

The document discusses the Simple Linear Regression Model, focusing on the relationship between the dependent variable (y) and the independent variable (x). It outlines key issues such as allowing for other factors affecting y, understanding the functional relationship, and ensuring a ceteris paribus condition. Additionally, it covers assumptions related to the error term (u) and the derivation of Ordinary Least Squares (OLS) estimates.

Uploaded by

Pavneet Singh
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/ 17

Econometrics

Pavneet Singh

IIM Amritsar

August 31, 2020

Pavneet Singh (IIM Amritsar) Session 3 August 31, 2020 1 / 10


Simple Linear Regression

Pavneet Singh (IIM Amritsar) Session 3 August 31, 2020 2 / 10


The Simple Linear Regression Model

Key Problem- Explain y in terms of x, where y is called the dependent


variable and x is called the independent/explanatory/control
variable/covariate
Three issues
1 How do we allow for other factors to affect y?
2 What is the functional relationship between y and x?
3 How can we be sure that we are capturing a ceteris paribus relationship
between y and x?

Pavneet Singh (IIM Amritsar) Session 3 August 31, 2020 3 / 10


The Simple Linear Regression Model

y = β0 + β1 x + u (1)

u is known as the error term, or the disturbance term

Pavneet Singh (IIM Amritsar) Session 3 August 31, 2020 4 / 10


The Simple Linear Regression Model

y = β0 + β1 x + u (1)

u is known as the error term, or the disturbance term

∆y = ∆β0 + ∆β1 x + ∆u (2)


If ∆u = 0 (keeping everything else constant), we get

∆y = β1 ∆x (3)
β0 is k/s the intercept parameter, β1 is k/s the slope parameter

Pavneet Singh (IIM Amritsar) Session 3 August 31, 2020 4 / 10


Some issues

Linearity implies that a one-unit change in x has the same effect on y,


regardless of the initial value of x
Is this always the correct assumption? If not, how to deal with it?

Pavneet Singh (IIM Amritsar) Session 3 August 31, 2020 5 / 10


Some issues

Linearity implies that a one-unit change in x has the same effect on y,


regardless of the initial value of x
Is this always the correct assumption? If not, how to deal with it?
The model is linear in parameters, not in x and y
√
y = β0 + β1 x + u is a linear regression model, but
y = 1/(β0 + β1 x) + u is not!

Pavneet Singh (IIM Amritsar) Session 3 August 31, 2020 5 / 10


Assumptions on u and x

Can we hold all the other factors in u fixed? No! We need some
assumptions about how x and u are related to each other

Pavneet Singh (IIM Amritsar) Session 3 August 31, 2020 6 / 10


Assumptions on u and x

Can we hold all the other factors in u fixed? No! We need some
assumptions about how x and u are related to each other
Note- u and x are random variables in this model!
A harmless assumption- E(u) = 0

Pavneet Singh (IIM Amritsar) Session 3 August 31, 2020 6 / 10


Assumptions on u and x

Can we hold all the other factors in u fixed? No! We need some
assumptions about how x and u are related to each other
Note- u and x are random variables in this model!
A harmless assumption- E(u) = 0
What about corr(u,x) = 0?, corr(u, x 2 ) = 0?
Mean independence

E(u|x) = E(u) (4)

Pavneet Singh (IIM Amritsar) Session 3 August 31, 2020 6 / 10


Assumptions on u and x

Can we hold all the other factors in u fixed? No! We need some
assumptions about how x and u are related to each other
Note- u and x are random variables in this model!
A harmless assumption- E(u) = 0
What about corr(u,x) = 0?, corr(u, x 2 ) = 0?
Mean independence

E(u|x) = E(u) (4)


Independence =⇒ Mean Independence =⇒ Corr(u,x)=0

E(u|x) = 0 =⇒ E(y |x) = β0 + β1 x (5)


The The Population Regression Function

Pavneet Singh (IIM Amritsar) Session 3 August 31, 2020 6 / 10


The Population Regression Function

y = E(y |x) + u
y as the sum of the systematic part explained by x, and the
unsystematic part

Pavneet Singh (IIM Amritsar) Session 3 August 31, 2020 7 / 10


The Regression line

Fitted Values, errors and residuals

Pavneet Singh (IIM Amritsar) Session 3 August 31, 2020 8 / 10


Deriving the OLS Estimates

We use the two assumptions discussed above, i.e., E(u|x) = E(u) = 0


to derive sample estimates
These assumptions also imply that cov(x,u) = E(xu) = 0

Pavneet Singh (IIM Amritsar) Session 3 August 31, 2020 9 / 10


Deriving the OLS Estimates

We use the two assumptions discussed above, i.e., E(u|x) = E(u) = 0


to derive sample estimates
These assumptions also imply that cov(x,u) = E(xu) = 0
A random sample, {(xi , yi ), i = 1, 2, ..., n}

yi = β0 + β1 xi + ui (6)

Pavneet Singh (IIM Amritsar) Session 3 August 31, 2020 9 / 10


Deriving the OLS Estimates

We use the two assumptions discussed above, i.e., E(u|x) = E(u) = 0


to derive sample estimates
These assumptions also imply that cov(x,u) = E(xu) = 0
A random sample, {(xi , yi ), i = 1, 2, ..., n}

yi = β0 + β1 xi + ui (6)

E(u) = 0 =⇒ E(y − β0 − β1 x) = 0 (7)

E(xu) = 0 =⇒ E[x(y − β0 − β1 x)] = 0 (8)

Pavneet Singh (IIM Amritsar) Session 3 August 31, 2020 9 / 10


Deriving the OLS Estimates

Solving for the sample counterparts of eqns 7 and 8 above


1 n
Σ (yi − βˆ0 − βˆ1 xi ) = 0 (9)
n i=1
1 n
Σ xi (yi − βˆ0 − βˆ1 xi ) = 0 (10)
n i=1

Pavneet Singh (IIM Amritsar) Session 3 August 31, 2020 10 / 10

You might also like