Lecture 3
1.Beginning to work with proper
econometrics software packages
EViews and PcGive
2. Estimators and their properties
Properties of least squares estimators
Maximum likelihood estimation
Monte Carlo studies
INEMET [U13783]
INEMET [U13783]
INEMET [U13783]
Unbiased estimation of 1
INEMET [U13783]
Gauss-Markov Theorem
Given the classical assumptions,
ordinary least squares (OLS) estimators,
in the class of unbiased linear estimators,
have the minimum variance.
INEMET [U13783]
Efficiency
The GaussMarkov theorem states that, provided that the regression model assumptions
are valid, the OLS estimators are BLUE: best (most efficient) linear (functions of the values
of Y) unbiased estimators of the parameters.
probability density
functions
OLS
other unbiased
estimator
1
INEMET [U13783]
27
probability density
functions
1
INEMET [U13783]
27
INEMET [U13783]
INEMET [U13783]
INEMET [U13783]
INEMET [U13783]
INEMET [U13783]
INEMET [U13783]
ui ~ N(0,1)
Considering small samples of say size n=30 we could fix a set of 30 X values.
Then using the random number generator of a suitable computer package
we could generate many sets of 30 values of ui
(say the number of replications R = 1000 or even 100000).
Next we use the equation to produce a set of Y values for each set of u values.
Now we can use OLS to regress each set of the Y values on X to obtain
1000 different estimates of the slope coefficient 1.
Now we can then check various claimed properties of OLS estimation such as
expectation = 0.5
variance = (1/x2)
We could also see what happened to the properties if we were, say, to use an
alternative distribution for the u values.
INEMET [U13783]
increased
INEMET [U13783]