Statalist


[Date Prev][Date Next][Thread Prev][Thread Next][Date index][Thread index]

Re: st: gsort issue


From   Jeph Herrin <junk@spandrel.net>
To   statalist@hsphsun2.harvard.edu
Subject   Re: st: gsort issue
Date   Thu, 05 Jul 2007 17:15:45 -0400

I'm with Fred & Nick. After using Stata exclusively for
over a decade, I still find that my intuition about how
Stata will treat a missing value to come short suprisingly
often, and this is another example of where it doesn't do
what I would think.

I've often wished for a "negative" missing (a la -99)
value, so I could simply:

   replace x = -abs(x) if missing(x)

so that in contexts where I would rather treat missing
as the smallest rather than largest value, I could do so
in one stroke instead of trying to trick, eg, -bys y (x)-
into doing what I want.


j





Nick Cox wrote:
I don't use -gsort- much, as I usually prefer to work out my own -sort- order without wanting to re-discover
the precise idiosyncratic syntax of -gsort-. (I've got a blind spot on -recode- for the same kind of reason.)
(That's not on a par with B*ll G???d, who can write the equivalent of an -egen- function several times faster than it takes to find out whether that function already exists.)
But -- to the point -- while what Brian says is a fair answer it seems to me to point to a missing option on -gsort-.
-reallydowantmissingfirst- would not be very Stataish as a name, but Fred Wolfe's want and need seemed very reasonable to me.
Nick n.j.cox@durham.ac.uk
Brian P. Poi

On Thu Jul  5 06:58:30 2007, Fred Wolfe wrote:

Is there a problem with gsort (Stata 10 and below) or am I misunderstanding something?

I have a variable called -phdif-. I want the greatest value of that variable to appear in the last observation. There are
missing values,
so I use -gsort- with the -mfirst- option.
...

. gsort phdif
. l phdif in 1,clean

phdif
1. 1

. l phdif in l,clean

phdif
169914. .

The problem appears to be that missings are still last even
though I
used the -mfirst- option.

Any suggestions? Is this a problem or am I thinking about this incorrectly?
The "mfirst" option of -gsort- applies only to variables sorted in descending order.

Stata stores missing values as extremely large numbers, so if a variable is sorted in descending order, missing values should appear first in the list since they are greater than all non-missing values.

-gsort-, however, tries to be helpful when sorting in descending order by putting the missing values at the end of the list, assuming that the user really cares about the large real values of the variable, not the missing values.

The "mfirst" option tells -gsort- to put the missing values first in the list instead of trying to be helpful by putting them at the end of the list.

If you want to get the missing values to appear first when doing an ascending sort, one way to proceed is to create a 0/1 variable equal to 0 if the variable of interest contains missing and 1 otherwise and then sort by the indicator variable and the variable of interest:

. sysuse auto
. generate missrep78 = cond(missing(rep78), 0, 1)
. gsort missrep78 rep78
. list rep78 in 1/7, sep(0)
+-------+
| rep78 |
|-------|
1. | . |
2. | . |
3. | . |
4. | . |
5. | . |
6. | 1 |
7. | 1 |
+-------+
*
*   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/


*
*   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/



© Copyright 1996–2014 StataCorp LP   |   Terms of use   |   Privacy   |   Contact us   |   What's new   |   Site index