[Gretl-users] Re: BVAR pkg
On 15.08.2023 13:18, Brian Revell wrote:
Hi Luca
a Bayesian version of multiple regression would also be very welcome
and much appreciated in the future
Dear Brian,
you can estimate Bayesian normal regression model via BMA package,
thought it is a little tricky. See the script below:
<hansl>
set verbose off
include BMA.gfn
open mrw.gdt --quiet
list Xs = nonoil intermed OECD gdpgrow inv school
smpl --no-missing gdp60 Xs
BMA(gdp60, Xs, _(Nrep=1, burn=0, alpha=0.999))
</hansl>
And now, what yoy get under 'Model: 1, posterior probability: 1.000000'
line is what you're looking for. Two remarks:
1. In fact 'alpha' should be set to 1.0, but it may generate errors (I
need to investigate it).
2. It looks like there is a bug in plotting graph where there are
missing values: to be fixed soon.
Marcin
--
Marcin Błażejowski