topic Re: Use label to subset my data in SAS Procedures

Use label to subset my data

rezausc — Thu, 14 Feb 2013 21:38:10 GMT

I have a dataset with over thousands of columns and millions of records. I don't need all the columns (need all the rows) and want to subset my dataset by keeping columns that have specific labels. So, the subsetting is not based on column or var names.... it's based on their labels. Any idea how to do it efficiently?

Thanks

Re: Use label to subset my data

ballardw — Thu, 14 Feb 2013 21:52:16 GMT

Do the labels of the variables you want to keep (or drop) have any unique characteristics such as a word or phrase(s) that do not occur in the labels of the other variables?

Or do you have an existing list of the labels you want to keep?

Re: Use label to subset my data

Reeza — Thu, 14 Feb 2013 21:52:59 GMT

Create a macro variable with the variable names by querying the dictionary.table with search for the specific label column.

proc sql;

select name into :vlist separated by " "

from dictionary.tables where libname ='SASHELP" and memname="CLASS" and upcase(label) contains "YOUR CRITERIA";

quit;

data want;

set have (keep= &vlist);

run;

Re: Use label to subset my data

rezausc — Thu, 14 Feb 2013 22:30:30 GMT

Yes. I'm looking for specific label (contains specific term). But the variables with this label are distributed across all vars.

Re: Use label to subset my data

rezausc — Thu, 14 Feb 2013 22:43:53 GMT

Sorry, but I'm a newbie to SAS... Not very familiar with dictionary.tables. When I ran your code, I got ths error:

ERROR: The following columns were not found in the contributing tables: name.

By the way, I'm running SAS in command line (no gui).

Re: Use label to subset my data

Reeza — Thu, 14 Feb 2013 22:49:57 GMT

Try googling dictionary tables sas and you'll find a lot of references to them.

SAS(R) 9.2 Language Reference: Concepts, Second Edition

Sorry, I pointed you to the wrong table. It is the sashelp.vcolumn or dictionary.columns table instead.

You can run a proc contents on this table to see the column names and even a proc print to see some of the data.

Check

proc sql;

select name into :vlist separated by " "

from dictionary.columns where libname ='SASHELP" and memname="CLASS" and upcase(label) contains "YOUR CRITERIA";

quit;

data want;

set have (keep= &vlist);

run;