topic Re: delete observations in SAS Procedures

delete observations

CharlesR — Wed, 14 Jul 2010 06:09:58 GMT

Hello, First time here. I have a data set where i would like to remove observations which equal a particular value of a variable. Here's what i've got:

data st.Def_2008;
delete all where (ResultType=77);
run;

Here's the error i get:
ERROR 79-322: Expecting a ;.

ERROR 76-322: Syntax error, statement will be ignored.

Help?

Re: delete observations

andreas_lds — Wed, 14 Jul 2010 06:29:43 GMT

You should read the "SAS 9.2 Language Reference: Concepts, Second Edition", especially the chapter about "DATA step concepts".

Link: http://support.sas.com/documentation/cdl/en/lrcon/62955/HTML/default/viewer.htm#/documentation/cdl/en/lrcon/62955/HTML/default/titlepage.htm

Re: delete observations

Ksharp — Wed, 14 Jul 2010 11:59:03 GMT

> data st.Def_2008;
> delete all where (ResultType=77);

Why not use 'not' operation symbol.

data st.Def_2008;
where ResultType ne 77 ;

Re: delete observations

sbb — Wed, 14 Jul 2010 12:17:44 GMT

There definitely needs to be more substance to the DATA step - either a SET (providing an input SAS member) or some type of assignment statements to derive one or more SAS variables and then use an IF THEN DELETE; or some similar-result oriented coding technique.

However, as was suggested, the SAS DOC and other supplemental technical reference material hosted at SAS.COM support website will be quite useful to the original poster.

Scott Barry
SBBWorks, Inc.

Re: delete observations

FredrikE — Wed, 14 Jul 2010 12:45:03 GMT

Try sql instead:

proc sql;
delete *
from st.Def_2008
where ResultType=77
;
quit;

Re: delete observations

CharlesR — Wed, 14 Jul 2010 16:37:27 GMT

Thanks so much for all the help everyone! Proc SQL seems like what I'm really looking for! I have multiple situations to eliminate observations and it's what is easiest for me to get a handle on.

Re: delete observations

SASKiwi — Wed, 14 Jul 2010 22:31:47 GMT

The equivalant DATA step logic is:

data st.Def_2008;
set st.Def_2008;
if ResultType=77 then delete;
run;

The advantage of the DATA step is you can do a lot more in the same step with only one pass of the data than you can in SQL, for example create more than one output dataset containing different results.

Re: delete observations

Ksharp — Thu, 15 Jul 2010 08:44:15 GMT

Thanks . Sbb.
There are more things i have to learn.And also forget SasKiwi's right code.:(

Re: delete observations

Anilsk — Thu, 15 Jul 2010 11:07:34 GMT

Just try this code:

data st.Def_2008; /*Destination dataset */
set st.Def_2008; /*Source dataset*/
if ResultType=77 then delete;

It should delete that particular value from the dataset.

Re: delete observations

sbb — Thu, 15 Jul 2010 14:25:33 GMT

Also consider using a WHERE statement, if appropriate, for efficiency.

Scott Barry
SBBWorks, Inc.

Re: delete observations

FredrikE — Fri, 16 Jul 2010 14:11:50 GMT

One advantage with proc sql is that index is kept and that you don't rewrite the table, useful when working with larga ds...