Hello there,
I have derived the following model,
w = f(x,y,z)
where x has a quadratic relation with w, while y and z are linearly related
to w.
I formulated the econometric model as follows,
w = a + b1x +b2x^2 + b3y + b4z + e
The variables are timeseries, subscripted from t to T, with a and e as the
constant and error terms respectively.
My questions are:
1. Have I committed a fundamental error (specification or otherwise) in
combining the linear and quadratic relations into a single regression model?
2. If I have, then what is the best alternative to test the relation between
w and x, y and z?
Thanks.
Noel
[log in to unmask]
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
|