topic Re: Delete observations between two dates and override dataset in SAS Programming

Delete observations between two dates and override dataset

jeremy4 — Tue, 15 Oct 2019 17:24:14 GMT

Hi,

I am looking to override an existing dataset ("test"), which deletes observations where the variable "YEAR_MONTH" is in the month of September 2018. Would it be possible to edit my code as the "delete" part of my code is not in blue text when I input my code, so there may be an issue?

Code

data test;
set test;
if '01Sep2018'd <= input(YEAR_MONTH,yymmn6.) <= '30Sep2018'd then delete;
run;

Re: Delete observations between two dates and override dataset

Jagadishkatam — Tue, 15 Oct 2019 17:29:01 GMT

Alternately use if not instead of delete

data test;
   set test;
   if not ('01Sep2018'd <= input(YEAR_MONTH,yymmn6.) <= '30Sep2018'd);
run;

Re: Delete observations between two dates and override dataset

novinosrin — Tue, 15 Oct 2019 17:30:17 GMT

Why reference the same dataset recursively. @ballardw once taught me never do that. The sage's advice has been well taken. Please use

Proc sql;

delete from table

where......................

quit;

Re: Delete observations between two dates and override dataset

hashman — Tue, 15 Oct 2019 19:46:10 GMT

@jeremy4:

First, as @novinosrin has said, there's no need to rewrite the entire file. Second, since you year_month variable is character in YYYYMM pattern, there's no need to convert it to a SAS date value. Just create a view like either of the two below:

data vtest / view = vtest ;                                             
  set have (where = (year_month ne "201809")) ;                         
run ;                                                                   
                                                                        
proc sql ;                                                              
  create view vtest as select * from have where year_month ne "201809" ;
quit ;

Then reference VTEST instead of TEST in your code downstream where the subset is needed.

Kind regards

Paul D.