Recommended maximum number of variables for data set in EG?

[deleted_user]

Is there a recommended maximum number of variables for a data set in EG? Processing a data set with about 500 variables and 1,000 observations produces these symptoms:

1 - The query window does not open the data set and displays a message "You can not create a query based on [data set name]."

2 - A PROC PRINT of the data set takes 9 minutes to execute. That is with a "where=" data set option that limits the input to about 50 observations.

Are these symptoms due to the number of variables? If they are, how many variables can EG handle? If they aren't, what other factors might be causing the symptoms?

Thanks,
Paulette

[ChrisHemedinger]

There is no inherent limitation on how "wide" your data should be to work within EG. Your first issue sounds like a problem that requires follow-up with tech support.

The second issue is probably due to volume and output format. 500 columns by 50 rows results in 25,000 table cells. If you are using the default output format of HTML, the resulting HTML result might be large and take a while to render. You can test this by setting your result options to use Text listing instead of HTML and see if the result appears faster.

One other note: Hotfix 5 for EG 4.1 addresses some performance problems with wide data. See http://support.sas.com/kb/19/738.html.

Here is a SAS program that generates a simple wide data set, for testing.

[pre]
%macro makewide;
data wide;
%do i=1 %to 700;
length var&i 8 char&i $ 20;
var&i = &i;
char&i = "Column &i";
%end;
do j=1 to 1000;
output;
end;
run;
%mend;

%makewide;
[/pre]

Chris

[Doc_Duke]

Paulette,

I've used both more rows and more columns successfully (though I've never tried to print an entire dataset that is that big!). This note shows some ways to get more efficient throughput for large data in EG

http://support.sas.com/kb/26/178.html

Doc