Econometrics
University of Milan-Bicocca
Course lecturer:
Maryam Ahmadi
[email protected]
1
Linear
Regression Model
2
Simple Regression Model
y= β0 + β1x+u
• Using a random sample of size n from the population: { 𝑥𝑖 , 𝑦𝑖 : 𝑖 = 1, … , 𝑛}, we can write:
yi = β0 + β1xi + ui
y= Variable being explained or predicted (dependent, explained)
x= Variable explaining or predicting y (independent, explanatory)
u,= factors other than x that affect y (error term or disturbance)
yi=ith observation of y, xi=ith observation of x
ui= ith observation of error term
β0 = Intercept coefficient
β1 = Slope coefficient
Example: wage= 𝛽0 + 𝛽1 𝑒𝑑𝑢𝑐 + 𝑢
❑ a single dependent variable and a single independent variable 3
Multiple Regression Model
More precise and reliable as the effect of education on wage, because the effects of experience and
gender are controlled for.
Intercept Slope parameters
Dependent variable,
explained variable, Error term,
Independent variables, disturbance,
response variable,…
explanatory variables, unobservables,…
regressors,…
4
➢We don’t know the population parameters βi.
→We need to estimate a sample regression function (SRF)
➢ Our estimate of the truth!
By applying estimation method on given data we estimate the numerical values of the parameters
(𝛽መi 𝑖𝑠 𝑡ℎ𝑒 𝑒𝑠𝑡𝑖𝑚𝑎𝑡𝑒𝑑 𝑣𝑎𝑙𝑒 𝑓𝑜𝑟 𝛽i )
5
Deriving the ordinary least squares (OLS) estimates
6
Deriving the ordinary least squares (OLS) estimates
• 𝑦𝑖 = 𝛽1 + 𝛽2 𝑥𝑖 +𝑢𝑖
• 𝑦ො𝑖 = 𝛽መ1 + 𝛽መ2 𝑥i
• 𝑢ො 𝑖 = 𝑦i − 𝑦ො𝑖
✓The smaller 𝑢ො 𝑖 , the better approximation is our line
The OLS estimators of the model parameters are those that minimise the sum of square of the residuals.
10
M𝑖𝑛 𝑢ො 𝑖 2
𝑖=1
7
OLS estimators for unknown regression parameters:
✓Simple regression model
[σ𝑛𝑖=1(𝑥𝑖 − 𝑥)(𝑦
ҧ 𝑖 − 𝑦)]
ത
✓ 𝛽መ1 =
σ𝑛𝑖=1 𝑥𝑖 − 𝑥ҧ 2
✓ 𝛽መ0 = ഥ𝑦 − 𝛽መ1 𝑥ҧ
𝑦ො is the Fitted Value or Predicted Value Of y
intercept, 𝛽0 , is the predicted value of 𝑦𝑖 when 𝑥𝑖 = 0
Regression Line: 𝑦ො𝑖 = 𝛽መ0 + 𝛽መ1 𝑥i ,
slope estimate 𝛽1 tells us how 𝑦ො changes when x changes by one unit.
Residual of the regression model∶ 𝑢ො 𝑖 = 𝑦i − 𝑦ො𝑖
8
Example 1
a) Write an econometric model in which the number of weeks participation in Training is the independent or
explanatory variable and Wage is the dependent variable.
Suppose we have the following data
Training represent the number of weeks participation in the funded training program and Wage represent
hourly wage in Euro.
b) How many observations do we have in the sample?
c) What are the averages of hourly wage and weeks participation in training in the sample?
d) Compute the slope coefficient and interpret it.
e) Compute intercept and interpret it.
9
Example 2
Consider a sample of N individuals. We observe wage rates (per hour), gender,
experience and years of schooling.
approximate wages by a linear model of the gender of the individual. And
interpret the results.
dummy: (= 1 if male, 0 if female).
wagei = β0 + β1 malei + ui
𝑤𝑎𝑔𝑒
ෟ = 5.15 + 1.16 malei
Intercept: when mail=0, that is, for females, predicted wage = $5.15
Predicted (or fitted) wage for males: 𝑤𝑎𝑔𝑒
ෟ = $5.15 +$1.16 = $6.31.
Slop parameter: if variable mail changes by one unit, the wage changes by $1.16.
The difference between the wage of males and females.
10
Summary
• 𝛽መ0 and 𝛽መ1 are ordinary least squares (OLS) estimates of 𝛽0 𝑎𝑛𝑑 𝛽1 .
• Using 𝛽መ0 and 𝛽መ1 , we define fitted (or predicted) value for y.
Fitted value (𝑦ො𝑖 ) is the estimate of 𝑦𝑖 , the value we predict for 𝑦𝑖 .
𝑦ො = 𝛽መ0 + 𝛽መ1 𝑥
➢The intercept, 𝛽መ0 , is predicted value of 𝑦𝑖 when 𝑥𝑖 =0.
➢The slope estimate 𝛽መ1 tells us how 𝑦ො changes when x changes by one unit.
11
Summary
➢The difference between the fitted value (𝑦ො𝑖 ) and the actual
observation (𝑦𝑖 ) is the residual (𝑢ො 𝑖 ): 𝑢ො 𝑖 =𝑦𝑖 -𝑦ො𝑖
➢It is important to know that: 𝑢𝑖 ≠𝑢ො 𝑖
Error terms residuals
Unobserved factors are known to us
not included in regressors
estimation errors (𝑦i − 𝑦ො𝑖 )
12
Problem
1- True or False? Explain.
• Error terms and residuals are the same concept.
2- Suppose that we are interested in the effect of income on consumption.
a) Determine the dependent and the independent variable and write the
simple regression model.
b) Suppose that we use a sample of 50 observations and by using OLS
estimator, we have 𝛽መ1 = 523 and 𝛽መ2 = 0.7. Write the OLS regression line.
c) How do you interpret the estimated parameters.
d) What is the fitted value of consumption if income is 10000.
Deadline: Wednesday 11th March, 2020, 24:00
13
