Re: [Gretl-devel] an Odd example with restrict -- NA as a standard error
On Sun, 18 Nov 2012, Lee Adkins quoted Allin:
> In this example the computed variance for the baths
> coefficient is 2.15e-13, which seems on the big side to be
> forced to zero. I think the only way to get this "right" would
> be to somehow keep track of which coefficients, if any, are
> assigned a definite numerical value by the restriction -- i.e.
> look for rows of the R matrix that have only one non-zero
> entry?
and wrote:
Yes, I think so, too since the std errors (and coefficients) can
always be
changed by an arbitrary amount by rescaling y or x or both.
I've now put that into CVS. Here's a precise account of what's new:
When transcribing from the numerically computed variance matrix for
restricted OLS estimates, we examine the R matrix (as in R\beta =
q). For each row of R, if it contains a single non-zero element we
record the column position of that element, call it p. And we
override the calculated values for all elements in row p and column
p of the variance matrix with zeros.
If anyone can think of a way this could end up doing the wrong
thing, please let me know!
Allin