[Date Prev][Date Next][Thread Prev][Thread Next][Date index][Thread index]
Re: st: random order after sub-sorting
At 6:03 PM +0200 8/12/05, D.C.J. Vissers wrote:
1. Why does Stata randomly order records after a certain sorting?
After issuing the following command:
any observations with equal values of varlist will be in random
order. That's just how -sort- works. There is an FAQ on this issue
(http://www.stata.com/support/faqs/lang/sort.html), and it is a
common gotcha even among experienced programmers. If instead you
want Stata to leave observations with tied values of varlist in the
same order as before, use the -stable- option (introduced in version
8, I believe).
2. How can I make sure that every time the same selection is made?
As the aforementioned FAQ describes, whenever you are selecting cases
based on their order, you should insure that that order is
unambiguous (read reproducible). In some cases, this can be
accomplished by adding additional variables to your -sort- command;
often, the choice of variable(s) is straightforward (e.g., time, or
perhaps in your case, status_of_wife). And in certain cases, it can
be helpful to create a variable (perhaps temporary) like this:
gen record_no = _n
immediately after you read in your data (e.g., using -use-,
-insheet-, or -infix-). This can then be used later on, either by
itself or in conjunction with other variables, whenever it is
necessary to establish a reproducible order among the observations.
* For searches and help try: