<>
Austin,
very good catch, thank you. Indeed, a -sort- can do a lot of damage,
and I should have left it out. While writing the reply I forgot that,
of course, -egen- is clever and will automatically create groups in
ascending order of v1, without altering the sort order of the data.
Therefore, one line is enought to solve Alexey's problem.
Eva
2009/4/17 Austin Nichols <[email protected]>:
> Eva--
> Good point--I am so used to using -egen-'s -group()- function for
> string variables, that solution did not occur to me. But the prior
> -sort- is superfluous, and changing the sort order can be dangerous
> for downstream code, when the current sort order contains useful
> information (hence my extra lines restoring current sort order).
>
> On Fri, Apr 17, 2009 at 4:59 PM, Eva Poen <[email protected]> wrote:
>> <>
>>
>> Alexey,
>>
>> what you want is not technically a rank. -egen, rank- will give you
>> ranks, taking all observations into account, and resolving ties. What
>> you want can be done with -egen- also:
>>
>> sort v1
>> egen rank = group(v1)
>>
>>
>> Hope this helps,
>> Eva
>>
>>
>>
>> 2009/4/17 Alexey Bessudnov <[email protected]>:
>>> Dear all,
>>>
>>> I have an interval variable v1 with values 1,1,15,15, 47, 47, etc. and I
>>> want to recode it into a rank variable v2 so that 1=1, 15=2, 47=3, etc. Is
>>> there any simple way to do this automatically? I have tried -egen rank,
>>> track-, but it returns v3.
>>>
>>> +------------+
>>> | v1 v2 v3 |
>>> |------------|
>>> 1. | 1 1 1 |
>>> 2. | 1 1 1 |
>>> 3. | 15 2 3 | 4. | 15 2 3 | 5. | 47 3 5 | 6. | 47 3 5
>>> |
>>> +------------+
>>>
>>> Many thanks!
>>>
>>> Alexey Bessudnov
>>> DPhil student in sociology
>>> St.Antony's College, Oxford University
>
*
* For searches and help try:
* http://www.stata.com/help.cgi?search
* http://www.stata.com/support/statalist/faq
* http://www.ats.ucla.edu/stat/stata/
