Title: The ten assumptions of the Gaussianstandardclassical linear regression model
1The ten assumptions of the Gaussian/standard/clas
sical linear regression model
- 1. A Population Regression Function (PRF) that is
linear in the parameters provides a good
explanation of observed variation in the data. In
other words a model of the following form
describes the data correctly.
2The ten assumptions of the Gaussian/standard/clas
sical linear regression model
- 2. Along the Population Regression Function
(PRF), the values of the independent variable, X,
are fixed in repeated sampling, i.e., X is
nonstochastic.
3The ten assumptions of the Gaussian/standard/clas
sical linear regression model
- 3.Along the Population Regression Function (PRF),
the distribution of actual Y values at each X
value is such that the distribution of the
deviations of actual Ys from the Y values at
each value of X along the PRF has a mean of 0. In
other words E(ui Xi ) 0.
4The ten assumptions of the Gaussian/standard/clas
sical linear regression model
- 4. Along the Population Regression Function
(PRF), the distribution of actual Y values at
each X value is such that the distribution of the
deviations of actual Ys from the Y values at
each value of X along the PRF has the same
variance at all values of X. In other words - Var(ui Xi) s2, or homoscedasticity is
required.
5The ten assumptions of the Gaussian/standard/clas
sical linear regression model
- 5. Along the Population Regression Function
(PRF), the distribution of actual Y values at
each X value is such that a deviation of an
actual Y value from the Y value on the PRF at
some value of X is not related to the deviation
at any other value of X. In other words Cov(ui,
uj Xi , Xj) 0, or no autocorrelation is
required.
6The ten assumptions of the Gaussian/standard/clas
sical linear regression model
- 6. Along the Population Regression Function
(PRF), the distribution of actual Y values at
each X value is such that a deviation of an
actual Y value from the Y value on the PRF at
some value of X is not related to the value of X.
In other words Cov(ui, Xi ) 0.
7The ten assumptions of the Gaussian/standard/clas
sical linear regression model
- 7. The number of observations is greater than the
number of variables or the number of parameters
to be estimated.
8The ten assumptions of the Gaussian/standard/clas
sical linear regression model
- 8. There is more than one value of X. In other
words Var(X) gt 0.
9The ten assumptions of the Gaussian/standard/clas
sical linear regression model
- 9. The model specification is correct (variables,
functional form, error characteristics).
10The ten assumptions of the Gaussian/standard/clas
sical linear regression model
- 10. There is not an exact relationship among the
independent variables. In other words, there is
no multicollinearity.