This is clearly programmable with your own program (e.g. defining an
-egen- function), but here is a more systematic way of approaching it,
just with canned code.
. sysuse auto, clear
(1978 Automobile Data)
. egen rank_lo = rank(mpg) , by(rep78) unique
. egen rank_hi = rank(-mpg) , by(rep78) unique
. egen lo2 = total(mpg / (rank_lo == 2)), by(rep78)
. egen hi2 = total(mpg / (rank_hi == 2)), by(rep78)
. egen min = min(mpg), by(rep78)
. egen max = max(mpg), by(rep78)
. tabdisp rep78, c(min lo2 hi2 max)
----------------------------------------------------------
Repair |
Record |
1978 | min lo2 hi2 max
----------+-----------------------------------------------
1 | 18 24 18 24
2 | 14 16 24 24
3 | 12 12 28 29
4 | 14 14 28 30
5 | 17 17 35 41
. | 14 19 26 26
----------------------------------------------------------
Note that the results for -rep78 == 1- are correct. There are just two
values, so the second largest is the minimum and the second smallest
is the maximum.
Nick
njcoxstata@gmail.com
On 11 April 2013 08:42, "Bartus Tamás" <tamas.bartus@uni-corvinus.hu> wrote:
> The code for, say, the lowest 5th value of variable x could be
>
> sort x
> gen tempvar = x in 5
> egen newvar = mean(tempvar)
>
> The code for the highest 5th value is
>
> gsort - x
> gen tempvar = x in 5
> egen newvar = mean(tempvar)
On 13/04/11, Caterina.ASTARITA@ec.europa.eu wrote:
>> is that possible to generate a new variable containing the nth lowest or nth highest value of another variable?
*
* For searches and help try:
* http://www.stata.com/help.cgi?search
* http://www.stata.com/support/faqs/resources/statalist-faq/
* http://www.ats.ucla.edu/stat/stata/
