Next: About this document ...
Variance of Residuals in Simple Linear Regression
Allen Back
Suppose we use the usual denominator
in defining the sample
variance and sample covariance for samples of size
:
Of course the correlation coefficient
is related to this covariance by
Then since
, it follows that
If we apply this to the usual simple linear regression setup, we
obtain:
Proposition: The sample variance of the residuals
in
a simple linear regression satisfies
where
is the sample variance of the original response variable.
Proof: The line of regression may be written as
where
. The residual of a point
is
, so:
Next: About this document ...
Allen Back
2007-11-01