On Sun, 15 Jan 2012, Henrique Andrade wrote:
Em 15 de janeiro de 2011 Allin escreveu:
(...)
>
> The original model has time-varying X = (const, INFL(-1),
> LHUR) and time-invariant X = LHUR(-1). Your revision has just
> one time-varying X (= LHUR) and no time-invariant terms, not
> even a constant. Is that really what you want?
>
Yes, it is.
OK, then the issue is that since your model has only one
parameter to be estimated (besides the variance), various
matrix quantities used in the original script become 1x1, and
these automatically become scalars. You need to force them to
be matrices. This applies to the lines
H = zeros(k,1) # original
matrix H = zeros(k,1) # revised
b0 = $coeff -> matrix b0 = $coeff
V0 = $vcv -> matrix V0 = $vcv
Allin Cottrell