For consistency with what?
Each program has its own design style.
The SAS/Stata difference on missing
values as very low/very high hinges
on the Stata preference for _not_ seeing
a wadge of missings first after
. sort myvar
. list myvar
Also, each program must make its own decisions
about missing data, in which a Murphy-type
Law applies:
Whatever you do is wrong for some circumstances.
By and large, Stata follows the Unix
tradition of assuming smart users and giving
minimal, or least restricted, output.
So, with -graph- or -regress- on variables with
some missing values they can't be used --
and there is no message reminding you of
that.
With -tabulate- the situation is not
so clear-cut.
However, with say
. tabulate row col, chi2
I personally would not want a default
in which missing categories became
extra rows and columns. To put
it another way, the difficulty
arises because -tabulate- has both
statistical and data management roles.
It seems to me on this Stata is, as
close as possible, being
consistent with its own general
conventions: you can't graph missings,
you can't do statistics with missings,
but you can do data management with
missings.
Nick
[email protected]
Joseph Coveney
>
> Whoops! Apparently not. (SAS codes missing values as the
> smallest possible
> value, while Stata codes them as the largest.)
>
> Perhaps, for consistency, Stata ought to include missing
> categories in
> -tabulate- by default . . .
*
* For searches and help try:
* http://www.stata.com/support/faqs/res/findit.html
* http://www.stata.com/support/statalist/faq
* http://www.ats.ucla.edu/stat/stata/
