Dear All,
I'm using the TSCSREG procedure in SAS to estimate a fixed/random effects model of the form:
y(t+1) = b0 + b1*x1(t) + ... + e(t)
The problem is that the some of the IVs are in fact, dependent on a) each other; and b) on the dependent variable. That is, e.g.,
xi(t)=f( xj(t) ) (a)
xi(t)=f( y(t) ) (b)
Is there a way to control for the endogeneity so that biased estimates will not occur? How is this done in TSCSREG or other SAS procedures (e.g. MIXED)?
Or do I have to find alternative IVs so as to circumvent the problem?
Thank you very much!