[Lapack] the standard errors of the parameter estimates

[Sven Hammarling]

Dear Wenhua,

Assuming that m > n, the sample variance s^2 is given by

    s^2 = r^T r/(m - n),

where r is the vector r = ( b_(n+1) b_(n+2) ... b_m )^T and b is the 
relevant column of the right hand side matrix B.

Hope that helps.  Best wishes,

Sven Hammarling.

P.S. Sorry to hear that you are switching from NAG!!