I don't know, but I can make a guess, knowing that if I am
wrong someone can improve on my guess. (According to Karl Popper,
this is the essence of scientific method.)
I guess this is purely historic. The intent of -cumul-, introduced
in Stata 3.0, was always, at least primarily, as a preprocessor to
allow plots of the cdf. For graphical purposes, -float- is surely
adequate.
I guess also that it has taken this long -- from 1992 to 2005 --
for someone to say that in fact precision does matter to them
here. (Perhaps others found the same thing, but just did the
calculations themselves from first principles.)
I guess also that StataCorp will oblige you and make the tweak in an
ado update sooner or later.
Nick
[email protected]
Jann, Ben
> the -cumul- command (which computes the empirical
> cumulative distribution function) enforces float
> storage type for the generated variable (that is,
> the storage type of the generated variable will
> always be float no matter what the default storage
> type is). This gives rise to precission problems
> in some applications.
>
> Can someone explain me the rationale behind
> enforcing float storage type in this context?
>
*
* 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/
