SYSTEM estimation: WLS vs. SUR
Posted: Tue Jul 15, 2014 2:26 pm
Hello. What are the differences between "Weighted Least Squares" method and "Seemingly Unrelated Regression" method?
I had a pool estimation with the following settings:
Estimation method = cross-section weights
Covariance method = cross-section SUR (PCSE)
Then, I specified the same models in a system object, because I wanted to yield the same results from a system estimation. What confused me was the fact that I had to choose the 'Estimation method' of "Weighted Least Squares (WLS)" instead of "Seemingly Unrelated Regression (SUR)" to replicate the results. I find that the manual refers to the "WLS" method as "Cross-equation weighting" method.
So, I'm confused. When I ran the pool estimation, I thought I chose SUR. However, the system estimation says that it was not SUR but weighted least squares? Please let me know why this happens.
Thank you very much.
Rakkoo
I had a pool estimation with the following settings:
Estimation method = cross-section weights
Covariance method = cross-section SUR (PCSE)
Then, I specified the same models in a system object, because I wanted to yield the same results from a system estimation. What confused me was the fact that I had to choose the 'Estimation method' of "Weighted Least Squares (WLS)" instead of "Seemingly Unrelated Regression (SUR)" to replicate the results. I find that the manual refers to the "WLS" method as "Cross-equation weighting" method.
So, I'm confused. When I ran the pool estimation, I thought I chose SUR. However, the system estimation says that it was not SUR but weighted least squares? Please let me know why this happens.
Thank you very much.
Rakkoo