Non-linear restriction on the intercept

[oksana]

Hi guys,

I need to run a regression: Y = -kB + BX + e, where Y and X are sampled variables and k and B are of interest. It looks like a linear regression BUT there is a non-linear restriction on the intercept and we can't simply divide the slope by the intercept to come up with the value of K. Do you have any idea as to how to estimate such an equation in EVeiws? Thanks a lot for your ideas!

[EViews Gareth]

Code: Select all

y=-c(1)*c(2)+c(1)*x

[oksana]

Thanks Gareth!! The only thing is that when estimating it using OLS, the result doesn't make any sense to me.... I wonder if it's actually appropriate to use OLS, given non-linear expression in the model. Thanks again!

[startz]

OLS is perfectly appropriate. However, if B is anywhere near zero the small sample properties of the test are very bad. A test for k=k0 with proper size, due to Fieller, is

Code: Select all

ls c x c(1)=-k0*c(2)