Roger Newson
> I am having a problem with the variable whose name is (e), which
appears to
> be generated whenever an estimation command is executed, and which
contains
> the results of the function -e(sample)-. This variable does not seem
to be
> well publicised anywhere accessible via -findit (e)-, but is listed
> whenever the user types
>
> list *
>
> or when the user types
>
> list *e*
>
> or even when the user types
>
> list ?e?
>
> Similarly, if the user types
>
> unab fifi:*
>
> or if the user types
>
> unab fifi:*e*
>
> or even if the user types
>
> unab fifi:?e?
>
> then the expanded varlist in the macro -fifi- includes the varname
(e).
>
> My problem is, how do you drop (e) from the data set without
dropping any
> other variables or estimation results? If I type
>
> ereturn clear
>
> then (e) ceases to exist, but so do the current estimation results.
If I type
>
> drop _all
>
> or if I type
>
> drop *e*
>
> or even if I type
>
> drop ?e?
>
> then Stata drops (e), and any other variables specified by the
abbreviated
> varlist provided by the user. However, if I type
>
> drop (e)
>
> then Stata replies
>
> variable e not found
> r(111):
>
> and refuses to drop (e). Is there a way of dropping (e) from the
data set
> in a Stata program without destroying anything else?
I don't have a solution to this. In fact I've never even
noticed (e) before. This raises the question of whether
it's been there for ages, or it's only recently become
visible as a result of some other change in Stata.
I can add that attempts to solve Roger's problem indirectly by
-keep-ing everything else also failed.
To follow Roger's speculation, it appears that (e) is just
e(sample) in another guise. It looks like a variable in many
ways, but it isn't a full-fledged variable; let's call it a
pseudovariable in ignorance of any unpublicised Stata Corp
terminology.
_n is another pseudovariable, one might say, although the
differences between it and (e) are just as crucial
as the differences. (For example, you can't drop _n
even indirectly; if you could, you would probably
break your Stata!)
What is a little more worrying is that there
seem to be small inconsistencies in how (e)
is treated.
. su
and
. su *
differ in whether (e) is included.
. d
and
. d *
don't, but
. ds
and
. ds *
do.... But then I have long assumed that
. su
. su *
. d
. d *
. ds
. ds *
. l
. l *
were all equivalent pairs, and typed the simpler
version every time.
I see two contradictory and equally plausible
stories here:
1 This is accidental, or even a mistake. Stata
Corp need to fix the code so that only their
outerwear is showing.
2 This is a subtle feature. Stata Corp have
introduced a nuanced distinction between
(every variable) and (every variable PLUS (e))
so that you can examine (e) whenever you wish,
so long as it exists.
Nick
[email protected]
P.S. at least one reason why you can't
look at (e) directly with e.g.
. l (e)
is that another rule, that varlists may
be parenthesised, leads to the parens being
ignored.
*
* 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/
