Re: [Gretl-devel] bug with missings, dummies (and panel?)
On Tue, 13 Apr 2010, Riccardo (Jack) Lucchetti wrote:
On Tue, 13 Apr 2010, Berend Hasselman wrote:
> I feel that The R and Octave result of NA for 0*NA is perfect reasonable
> (in a numeric context). I prefer it.
Matter of taste.
Actually, I think it's a bit more than a matter of taste. If a
program evaluates 0*NA as NA I submit that shows that the program
does not implement the concept of "NA" as missing value.
The original source of "true" NAs in econometrics is observations
that were not made, were lost, or are otherwise unavailable. In
that context, as Jack said, "NA" is a place-holder for an unknown
value, but if it's the unknown value of a measurable quantity it
cannot be infinity or "Not a Number" (which is a place-holder for
a mathematical error), and multiplication by zero yields zero for
any such unknown but in principle valid value.
If a program has 0*NA = NA it is treating NAs as if they were
nans, which in general is not correct (though is is convenient for
the coder not to have to make the distinction).
Allin