Fit y = a + bx to a set of data 
. We choose a and b
such that 
. This is
the method of least squares and is a special case of the method of
maximum likelihood.
Essentially we are minimising the 
statistic. We find that:
.