Lecture notes 9 Non normality Flashcards
1
Q
What is non-normality
A
In the CLRM we assume that the regression error has a normal distribution.
However, in economics sometimes we do not have continuous variables and can have discrete ones so this error does not fit.
2
Q
What is a test for normality?
How is each part of this test done?
A
Jacques- Bera test
n/6[S^2 + (K-3)^2/4]
if this is close to 0 we conclude epsilon is normal.
S = skewness
S = M3/ M2^3/2
K = M4/ M2^2
3
Q
What is known about the standard normal distribution about normality?
A
If you divide normal distribution by epsilon squared it yields (0,1)
In standard normal 3rd moment (Skewness = 0)
4th Moment Kurtosis = 3
4
Q
How do you work out m2, m3, m4
A
m2 1/n sum of i=1 to n e^2
m3 1/n sum of i = 1 to n e^3
m4 1/n sum of i =1 to n e^4
