- Mark as New
- Bookmark
- Subscribe
- Mute
- RSS Feed
- Permalink
- Report Inappropriate Content
Posted 05-11-2011 07:58 AM
(886 views)
Hi,
This is the regression model i'm using:
proc reg data=data_regressie4 outest=regressie4 edf;
BY ANALYS;
model geschaalde_FE = geschaalde_deviation age size total_experience firm_experience aantalsector aantalbedrijf;
run;
The problem is that the estimate for geschaalde_deviation should be greater than 0. Is there a way to make sas estimate this model with "geschaalde_deviation" being greater than 0?
Thanks a lot!
This is the regression model i'm using:
proc reg data=data_regressie4 outest=regressie4 edf;
BY ANALYS;
model geschaalde_FE = geschaalde_deviation age size total_experience firm_experience aantalsector aantalbedrijf;
run;
The problem is that the estimate for geschaalde_deviation should be greater than 0. Is there a way to make sas estimate this model with "geschaalde_deviation" being greater than 0?
Thanks a lot!
1 REPLY 1
- Mark as New
- Bookmark
- Subscribe
- Mute
- RSS Feed
- Permalink
- Report Inappropriate Content
There is a RESTRICT statement in PROC REG, but it doesn't allow "greater than" restrictions, just equality restrictions.
The problem isn't really SAS. The problem is that when you perform a regression with multiple correlated variables (you have 7 variables in your model), least squares regression (not SAS) can have huge variances for your parameter estimates, and because of that, you can get a negative estimate even when you think it should be positive. In this situation, interpreting the coefficients is a meaningless exercise.
Sometimes using Partial Least Squares regression (PROC PLS) can alleviate this problem, but your R-squared from PLS will not be as good as from PROC REG, and there's no guarantee that PLS will provide positive estimates.
The problem isn't really SAS. The problem is that when you perform a regression with multiple correlated variables (you have 7 variables in your model), least squares regression (not SAS) can have huge variances for your parameter estimates, and because of that, you can get a negative estimate even when you think it should be positive. In this situation, interpreting the coefficients is a meaningless exercise.
Sometimes using Partial Least Squares regression (PROC PLS) can alleviate this problem, but your R-squared from PLS will not be as good as from PROC REG, and there's no guarantee that PLS will provide positive estimates.