turn on suggestions

Auto-suggest helps you quickly narrow down your search results by suggesting possible matches as you type.

Showing results for

Find a Community

- Home
- /
- SAS Programming
- /
- SAS Procedures
- /
- Need help with a clustering problem

Topic Options

- Subscribe to RSS Feed
- Mark Topic as New
- Mark Topic as Read
- Float this Topic for Current User
- Bookmark
- Subscribe
- Printer Friendly Page

- Mark as New
- Bookmark
- Subscribe
- Subscribe to RSS Feed
- Highlight
- Email to a Friend
- Report Inappropriate Content

01-12-2014 06:09 PM

I'd like to run a regression where the dependent variable (say, stocks' abnormal returns) are related to a few explanatory variables. The problem is that all the stocks come from the same industry,and so the error terms are probably correlated. Is there a SAS function that can take care of this problem? Pleade note that the data are NOT time-series data. The abnormal returns are measured at one single point in time. In other words, the data are just cross-sectional data. Thanks.

- Mark as New
- Bookmark
- Subscribe
- Subscribe to RSS Feed
- Highlight
- Email to a Friend
- Report Inappropriate Content

Posted in reply to Everton

01-12-2014 07:29 PM

Your best bet is likely to either use GEE (Generalized Estimating Equations) in Proc Genmod SAS/STAT(R) 9.22 User's Guide or here are another couple of ways to use clusters in SAS http://nesug.org/proceedings/nesug06/an/da01.pdf Hope this helps.

- Mark as New
- Bookmark
- Subscribe
- Subscribe to RSS Feed
- Highlight
- Email to a Friend
- Report Inappropriate Content

Posted in reply to overmar

01-18-2014 08:42 PM

Thanks, overmar. I've tried to learn how to use Proc Genmod and also Proc Surveyreg. But I don't think they can be used in my case. My sample all comes from the same industry, and so there is only class of observations (i.e., no multiple clusters). I just want to run a regression on this sample. Since the error terms may be correlated, I'd like to know if there a SAS function that can adjust the standard errors of the coefficients to account for the correlation. Thanks.

- Mark as New
- Bookmark
- Subscribe
- Subscribe to RSS Feed
- Highlight
- Email to a Friend
- Report Inappropriate Content

Posted in reply to Everton

01-18-2014 10:27 PM

You're looking for a way to adjust for dependent data.

I don't believe there's an easy way. One is to cluster the data, if you know what the clusters are, but since you have only one cluster I'm not sure. More likely it seems like its simply a limitation of your analysis rather than something you can control for in any way.

- Mark as New
- Bookmark
- Subscribe
- Subscribe to RSS Feed
- Highlight
- Email to a Friend
- Report Inappropriate Content

Posted in reply to Reeza

02-08-2014 12:07 PM

Thanks, Reeza. Sorry I didn't thank you earlier. I got a flu that knocked me out during the past two weeks.