PROC SORT

[JasonNC]

Hi Folks,

'when we are using options like noduprecs or nodupkey with Proc Sort does the dataset need to be sorted before

For ex:

1 menthod

Proc sort data= stats;

by state region;

run;

proc sort data= stats (noduprecs/nodupkey);

by state region;

run;

or

2 methos

Proc sort data=stats (noduprecs/nodupkey);

by state region;

run;

Is there any problem with 2 method.I think it will remove duplicate observations.

[art297]

: First, I have NEVER seen a situation where NODUPRECs is of any value.  If you don't want duplicates, use NODUPKEY and specify the by variables that you want to ensure aren't duplicated.

Second, no, you don't have to presort your data unless you want to control which of the duplicate records gets deleted.  E.g., if the records included a date field and you only wanted to keep the most recent date, you might use two sorts like:

proc sort data= stats;

  by state region descending date;

run;

proc sort data= stats nodupkey;

  by state region;

run;

[Reeza]

Be aware of what noduprecs is doing versus nodupkey, there are separate options.

http://www2.sas.com/proceedings/sugi30/037-30.pdf