topic Re: How to drop the extremum in SAS Data Management

How to drop the extremum

Xinhui — Wed, 18 Oct 2017 23:53:18 GMT

Let's suppose I have 100 values in my data set. If I want to drop the three highest values, as well as the three lowest values, what code would best accomplish this?

Re: How to drop the extremum

WarrenKuhfeld — Thu, 19 Oct 2017 00:14:57 GMT

data x;
do i = 1 to 100;
   r = uniform(368);
output;
end;
run;

proc sort data=x; by r; run;

data x2;
   set x;
   if _n_ gt 3 and _n_ lt 98;
run;

Re: How to drop the extremum

Reeza — Thu, 19 Oct 2017 01:42:54 GMT

How are you planning to deal with duplicates. If you have 5 values of 100 as the highest, do all 5 go? What about the next 3 99s?

The rules are ambiguous.

Re: How to drop the extremum

Xinhui — Thu, 19 Oct 2017 02:54:56 GMT

First, thank you for your valuable reminder.

Since I want to get rid the influence of extremum on my standard deviation, I think I should consider duplicate numbers as one value. So, in your example, if I have 5 duplicates of 100, they would all be gone. They are considered one value, not five.

Re: How to drop the extremum

Reeza — Thu, 19 Oct 2017 03:57:03 GMT

If that’s the goal look at the TRIMMED and/or WINDSOR option in PROC UNIVARIATE.