|
Paul ,
O there are so many ways to get this, but lets assum you want all of the
data sets in your work library and you want to use SQL:
Data D1 ;
Do I = 1 To 250 ;
Output ;
End ;
Run ;
Data D2 ;
Do I = 1 To 450 ;
Output ;
End ;
Run ;
Data D3 ;
Do I = 1 To 850 ;
Output ;
End ;
Run ;
Proc SQL ;
Select MEMNAME Label = 'DataSet' , NOBS Label = 'NOBS'
From Dictionary.Tables
Where Libname = 'WORK' ;
Quit ;
Toby Dunn
From: Paul Walker <walker.627@OSU.EDU>
Reply-To: Paul Walker <walker.627@OSU.EDU>
To: SAS-L@LISTSERV.UGA.EDU
Subject: Create Table Showing Number of Observations in Each Dataset
Date: Tue, 18 Jul 2006 18:50:42 -0400
I would like to create a table that looks like the following:
dataset nobs
------- ---------
D1 10000
D2 12000
D3 7500
The column 'dataset' represents the name of a dataset, and nobs is the
number of observations. Suppose there are 3 datasets as indicated, D1,
D2, and D3. There are several ways to create such a table using macro
variables or dictionary tables. However, I would like to use SQL to build
this table in the following sort of way.
proc sql noprint;
create table work.nobs_in_datasets as
union a (count(*) as nobs, 'D1' as dataset from work.D1)
b (count(*) as nobs, 'D2' as dataset from work.D2)
c (count(*) as nobs, 'D3' as dataset from work.D3);
quit;
However, I know the above code does not work and wondered if any PROC SQL
experts could help correct my syntax? The basic problem is how to 'stack'
the rows of data created by the code:
count(*) as nobs, 'D1' as dataset from work.D1;
count(*) as nobs, 'D2' as dataset from work.D2;
count(*) as nobs, 'D3' as dataset from work.D3;
TIA,
-PW
| 
