Beruflich Dokumente
Kultur Dokumente
Topic 7
Heteroscedasticity
1
Summary
2
The Nature of Heteroscedasticity
Heteroscedasticity is a systematic pattern in the errors
where the variances of the errors are not constant.
3
Regression Model
Yi = 1 + 2Xi + ui
heteroskedasticity: var(ui) = i2
4
Homoscedastic pattern of errors
consumption
Yi
.
. . . .
.. . . . .
. . . ...
... .. .
. . .. . .
. .. . . . .
..
. .
.
income Xi 5
The Homoscedastic Case
f(Yi) Yi
on
i
pt
u m
s
co
n
.
.
.
.
X1 X2 X3 X4 Xi
income 6
Heteroscedastic pattern of errors
consumption
.
Yi .
. .
. . . .
. .
. . . .
. . . . .
. . . .. . . .
. . . . . . .
. . . . .
. . . . .
. .
income Xi 7
The Heteroscedastic Case
f(Yi)
Y
i
on
p ti
u m
s
.
on
c
. rich people
.
poor people
x1 x2 x3 income Xi 8
Sources of Heteroscedasticity
9
Consequances of Heteroscedasticity
11
Detecting Heteroscedasticity
1. Park test
2. Glejser test
3. Whites general hetroscedasticity test
12
Detecting Heteroscedasticity: Park test
13
Whites general hetroscedasticity test
14
Detecting Heteroscedasticity
4. Other tests
a) Spearmans rank correlation test
b) Goldfeld-Quandt test
c) Bartletts homogeniety of variable test
d) Peak test
e) Breusch-Pagan-Godfrey test
f) CUSUMSQ test
15
Remedial Measures
16
Proportional Heteroscedasticity
Yi = 1 + 2Xi + ui
The variance is
assumed to be
where i2 = 2 Xi proportional to
the value of Xi
17
std.dev. proportional to Xi
Yi = 1 + 2Xi + ui
variance: var(ui) = i2 i2 = 2 Xi
standard deviation: i = Xi
var(u*i ) = 2
ui is heteroscedastic, but u*i is homoscedastic.19
Generalized Least Squares
These steps describe weighted least squares:
1. Decide which variable is proportional to the
heteroscedasticity (Xi in previous example).