On Sun, 7 Feb 2016, Igor Pozdeev wrote:
Hi all,
does the Nadaraya-Watson estimator work for everyone when run from the
interface? On my machine, when I click on "OK" (as shown below) nothing
happens (literally nothing). Which was not he case, by the way, before I
updated to Jan2016.
Thanks for the report. The "glue" linking the argument selector to the
code that actually calls the nadarwat function had come unstuck. This
is now fixed in git; snapshots will follow.
Also, it seems impossible to use a lagged variable as a regressor:
as shown below, lagged guys do not appear at all, and explicitly
writing "r_fx(-1)" or "r_fx1" does not help (did not help before the
update too).
I think this is an artifact arising from the fact that the
Nadaraya-Watson estimator is implemented as a function rather than as
a "command" (and so has a special variant of the argument selector).
This should be fixable too; I'll work on that today.
Allin Cottrell