Dear Clyde,
I believe it's just a convention. There was a choice to abort with an
error or to return a missing as a result. The second was chosen and (I
believe) there would have been an equal amount of disappointment if
the first one was chosen, e.g. "why can't Stata simply return a
missing and continue executing the program if no labels are defined?".
A more frequent, but very similar problem is something like:
local position_index=`positon_index'+1
which takes hell of a time to find, especially if it's located in a
rarely executed brunch of the program. I was told there is no way this
is going to change because too many programs rely on this behaviour
and there is nothing can be done within the current versions.
Note that in Mata there is:
"mata set matastrict sets whether declarations can be omitted inside
the body of a program. The default is off. If matastrict is switched
on, compiling programs that omit the declarations will result in a
compile-time error;"
but there is no equivalent for .ado programs.
Best regards,
Sergiy Radyakin
On 7/17/08, Clyde Schechter <[email protected]> wrote:
> This isn't a question--it's a warning about something that bit me, and a
> hope that Stata will change it in the future. (Or perhaps somebody will
> explain why it's actually a useful feature.)
>
> The following command:
>
> gen obs_interval = last_follow_up - first_visit if live_status ==
> "Alive":live_statusw
>
> resulted in obs_interval being set to missing for all observations because
> there is no such value label in the dataset as live_statusw. live_statusw
> was a typographical error--I meant to type live_status. (The reason the
> problem wasn't obvious immediately is that subsequent replace obs_interval
> = statements led to most records having a value for obs_interval, so that
> the incorrectly missing values didn't make things go clearly wrong until
> many analyses later.)
>
> Why doesn't Stata cease execution and complain that there is no such value
> label, just as it does if I try to use a non-existent variable name. (OK,
> Stata will allow abbreviated variable names, but "live_statusw" is not the
> abbreviation of any existing value label in the data set either.) Is
> there some purpose to having Stata allow reference to non-existent value
> labels? (I'm referring to the non-existence of the value label set
> live_statusw here. I understand why a reference to "xxx":live_status
> should not lead to an error even if live_status does not include "xxx"
> among its possible values.)
>
> Clyde Schechter
> Dept. of Family & Social Medicine
> Albert Einstein College of Medicine
> Bronx, NY, USA
>
>
>
> *
> * For searches and help try:
> * http://www.stata.com/help.cgi?search
> * http://www.stata.com/support/statalist/faq
> * http://www.ats.ucla.edu/stat/stata/
>
*
* For searches and help try:
* http://www.stata.com/help.cgi?search
* http://www.stata.com/support/statalist/faq
* http://www.ats.ucla.edu/stat/stata/
