Location via proxy:   [ UP ]  
[Report a bug]   [Manage cookies]                

Session 2 - 25.10.23

Download as pdf or txt
Download as pdf or txt
You are on page 1of 20

Econometrics I

EPOS Study Group

Session 2: 25/10/2023
Outline
I. Solutions for Last Week’s Exercises
II. Week’s Content
III. Exercises
IV. Q&A
Solutions of Last Week’s Exercises
1. Calculate the determinant and the rank of the following matrix:
Solutions of Last Week’s Exercises
1. Calculate the determinant and the rank of the following matrix:

Since det(A)=0  matrix A is not of


full rank
Solutions of Last Week’s Exercises
1. Calculate the determinant and the rank of the following matrix:

The maximum rank that matrix A can


have is 3, since min (3,3) is 3!

As rank(A) = 2 < 3  Matrix A is not of


full rank
Solutions of Last Week’s Exercises
1. Calculate the determinant and the rank of the following matrix:

Method 2: Echelon form of the matrix!


:

The maximum rank that matrix A can


have is 3, since min (3,3) is 3!

As rank(A) = 2 < 3  Matrix A is not of


full rank
Solutions of Last Week’s Exercises
2. Calculate the inverse of the following matrix:

:
Solutions of Last Week’s Exercises
Solutions of Last Week’s Exercises
OLS Estimator
• Set up: data on two random variables x and y (e.g. x is the
dollar value of investment in advertising in each company and y
is the dollar value of sales of each company)
• Objective: to establish a linear relationship between the two
variables allowing informed decisions to be made (e.g.
companies want to know how much investing in advertising
translates into sales so they can determine the optimal amount
to invest)
• Problem: it is impossible to obtain data for every existing
company (i.e. the population)  an estimation method is
required
OLS Estimator
OLS Estimator
• Ordinary Least Squares: the regression line is obtained such
that the sum of the squared residuals is minimized (i.e. the
smallest possible)
• In scalar form:

• In matrix form:

• Resulting estimator: bOLS = (X’X)-1X’y


Note: some residuals are above the line (positive) and others are below the line (negative)
 we use the squared residuals so that they do not cancel each other out
OLS Estimator
min e’e = (y-Xβ)’(y-Xβ) since y = Xβ+e
= y’y – y’Xβ – β’X’y + β’X’Xβ
= y’y - 2β’X’y + β’X’Xβ = g(β)

FOC = ∂g(β) / ∂β =! 0
-2X’y + 2X’Xb =! 0
2X’Xb = 2 X’y
X’Xb = X’y
b = (X’X)-1X’y
OLS Estimator

(0, σ2)
OLS Estimator
OLS Estimator
• Interpretation: the slope of the regression line indicates the relationship between the
analyzed variables (e.g. investing one extra dollar on advertising increases sales by 0.8
units)
 The expected relationship is marginal and each coefficient must be interpreted
separately (e.g. on average, investing one extra dollar on advertising increases sales by 0.8
units c.p.)
OLS Estimator
• Interpretation: depends on the units of the variables
Exercise (solution next week)
Exercise (solution next week)
1. Write the estimation model in matrix form, specifying each matrix with its elements and
dimensions.
2. What are the expected signs for the coefficients 𝛽2 and 𝛽3?
3. Calculate the OLS estimator b and give an interpretation for each coefficient.
4. How would the predicted number of sold cartons of orange juice be, if the price for one
carton is 80 cents and the expenditure on advertisement is 150,000 euros?
Questions?...

You might also like