Date:         Mon, 18 Feb 2008 09:35:40 -0800
Reply-To:     webonomic <webonomic@GMAIL.COM>
Sender:       "SAS(r) Discussion" <SAS-L@LISTSERV.UGA.EDU>
From:         webonomic <webonomic@GMAIL.COM>
Organization: http://groups.google.com
Subject:      Keeping unique variable while reshaping dataset.
Comments: To: sas-l@uga.edu
Content-Type: text/plain; charset=ISO-8859-1

The following is a representation of my dataset of survey responses
DATA surveys;
     INPUT surveyID respondent code;
CARDS ;
1       1       16
1       .        7
1       .        100
1       .        100
2       1       100
2       .        8
2       .        100
3       1       100
4       1       6
4       .        9
;
run;

There are a total of 4 survey respondents.  Each respondent can have 1
or more codes.  The respondent variable is given a value of 1 only
once for each ID.

I need 2 new datasets.  One is where code=100 and the other where code
^=100.  Datasteps with only these where clauses produce incorrect
datasets because I end up losing the correct representation of
respondents.

Where Code^=100 would produce a dataset such that the sum of
respondents=2 but it should be 3, because there are 3 unique
surveyID's.
surveyID        respondent      code
1                       1                       16
1                       .                       7
2                       .                       8
4                       1                       6
4                       .                       9

The resulting dataset I want is:
surveyID        respondent      code
1                       1                       16
1                       .                       7
2                       1                       8
4                       1                       6
4                       .                       9

A similar issue occurs for code=100.  I could count unique surveyID's
to get the correct number of respondents, but I have many other SAS
programs dependent on the input of this dataset in this format.

When respondent=1, code may or may not equal 100.

One way to do it would be to assign a 1 to the first 100 for each
surveyID then I think my Where clauses would work.

Thank you very much for any help you can give me.