If b(W) is linear then the model is just a linear
regression. If b(W) is linear and includes error term then
the model is a linear regression with random coefficients.
Regression with random coefficient is a linear regression
with heteroskedasticity of known form and could be estimated
using maximum likelihood method or generalized method of
moments.
Very useful treatment of linear/nonlinear regression,
regression with heteroskedasticity, maximum likelihood and
generalized method of moments could be found in
R. Davidson & J.MacKinnon, Estimation and Inferense in
Econometrics
-----------------------------
Alexander Tsyplakov
Novosibirsk State University
http://www.nsu.ru/ef/tsy/
Vadim Marmer wrote...
> Does anybody know if there is literature that discusses
the following type
> of regression models:
>
> Y=X*b(W)+u
>
> i.e. linear regression, but coefficients are a function
(possibly random)
> of some variable W. I am interested in estimation
(parametric or
> non-parametric) of this kind of model.
=================================================================
Instructions for joining and leaving this list and remarks about
the problem of INAPPROPRIATE MESSAGES are available at
http://jse.stat.ncsu.edu/
=================================================================
