As per the code, in the compress function used compress(x, 'ABCD', 'L'), we are suggesting compress to remove "ABCD" and any lower case letters, so even though the ABCD are all in caps, since we have mentioned them in compress function they are getting removed, in case you wish to remove only the lower case letters then try compress(x,, 'L'), this will remove only the lower case letters the output will be 123-4567-8901 B 234-5678-9012 /*code*/ data test; x='123-4567-8901 B 234-5678-9012 c'; y=compress(x,,'L'); proc print; run; Thanks, Jagadish ... View more

Another approach of using arrays to get the means of the variables is as below array avg(*) y1-y1000; average=mean(of avg(*)); Thanks, jagadish ... View more

if we consider that there are variables like below y1 y2 y3..............y1000 then we can take the average of all these variable in the below way average=mean(of y1-y1000); This way we will get the average of all the 1000 variables into one variable. Thanks, Jagadish ... View more

Yes Tom, the sas is unable to recognize date in 'yymmdd'd literals, so the best way would be to read the date with the informat yymmdd10. and use the date9. format as rightly mentioned by you. ... View more

The main reason for suggesting the use of the code "if date_variable>=input('2013-07-26',yymmdd10.) ;", as per the query the date when used this way '2013-07-26'd is not recognized by the system and is producing the errors as below ERROR: Invalid date/time/datetime constant '2013-07-26'd. ERROR 77-185: Invalid number conversion on '2013-07-26'd. The system is unable to recognize the date given in the format '2013-07-26'd, so it is essential that we read the date using the input function and then pass the condition. This code "if date_variable>='26jul2013'd;" will work well. However i have given an alternative way to this. Both the way will definitely work. i am open to any further suggestions. Thanks, Jagadish ... View more

Is it possible for you to provide the sample data and the output you wish to get to provide better answer. ... View more

Since you want any character value if present to be true you can try the below code Proc sql; select var1,var2, case when var3 ne ' ' then case when var4>10 then 'y' else 'n' end else 'n' end as flag from have; quit; Hope this works Thanks, Jagadish ... View more

Since both the format and the informat are yymmdd10. for the date, make sure that the date_variables are also in the same format and informat, then you can use the following code if date_variable>=input('2013-07-26',yymmdd10.) ; Hope this helps. Thanks, Jagadish ... View more

Hi, Please check the below code, its is producing the accurate results for time as well proc sql;     create table text_ as select date, case when length(strip(date))=16 then input(date||':00',is8601dt.) when length(strip(date))=10 then input(compress(date||'T00:00:00'),is8601dt.) else . end as tmp_var format=datetime20. from text; quit; output;                         Obs date                         tmp_var                         1 2013-02-02T08:00  02FEB2013:08:00:00                         2 2013-02-01        01FEB2013:00:00:00 Thanks, Jagadish ... View more