topic Re: Proc Robustreg in Statistical Procedures

Proc Robustreg

deleted_user — Wed, 31 Mar 2010 16:25:01 GMT

Hello and great forum!!

I have undertaken a fixed income portfolio analysis for one of my University's endowment funds, and I am working on a model to calculate returns. I have constructed a multiple linear equation and know one of the regressor coefficients. Does anyone know of a way for me to hold the known value constant, while letting sas adjust the coefficients on the other independent variables?

Any help or direction is greatly appreciated!

Tim

Re: Proc Robustreg

Paige — Wed, 31 Mar 2010 19:45:15 GMT

ROBUSTREG wasn't designed for this situation.

A possible way around this would be to fit the model with the known coefficient, and then compute residuals and do ROBUSTREG on the residuals. I haven't thought this through 100% to convince myself that is reasonable, but it sounds like it would work at first glance.

Re: Proc Robustreg

Dale — Thu, 01 Apr 2010 19:15:06 GMT

If you have the model

Y = b0 + b1*X1 + b2*X2 + ... + bk*Xk + C*Z

where C is known a priori, then you can rewrite the equation as:

Y* = Y - C*Z = b0 + b1*X1 + b2*X2 + ... + bk*Xk

So, all you need to do is to construct a new response variable in which you subtract the known effect from the original response. You can then fit your model employing this new response and predictors X1, X2, ..., Xk.

Re: Proc Robustreg

deleted_user — Fri, 02 Apr 2010 04:37:26 GMT

Thank you, Paige and Dale!

My data now looks great!

Tim