Good question.
Similarly, why does -max(1, .)- equal 1 (and not missing)?
But you might as well ask why -regress- doesn't
report that a regression can't be defined if there are _some_ missing
data in what you ask it to look at.
The larger point is that there are contexts in which missing data are
ignored (and everyone agrees that this is what they want) -- and that
applies to most statistical commands -- and at another extreme contexts
in which missing data just cannot be ignored -- the canonical example
for me being -sort-, in that observations with missing values _must_ go
somewhere.
So, the questions are
(a) where precisely in the middle do you draw the line in terms of
ignoring or including missings?
(b) what if any warning messages do you expect Stata to give?
An interesting feature of this thread to me -- and this issue has arisen
frequently over the life of the list -- is the emphasis many users are
now placing on (b) rather than (a).
On (a), list members can test themselves. Suppose x is missing on some
observations. What do you consider appropriate Stata behaviour for
. list x if x > 42
. regress z y if x > 42
That is the (friendly) battleground as I see it.
Nick
[email protected]
Simon Moore
I'm a bit confused. Why is that '.' has a value in gen, egen and
replace but the maximum value returned by su is not '.'? Surely this is
inconsistent?
*
* 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/