LSTAR model

[cychang]

hi everyone,

I would like to estimate the following LSTAR model:

yt = a0 + a1yt-1 + (b0+b1yt-1) / (1+exp(-gamma*yt-1) + et

and my code is:

equation eq1.ls y=c(1)+c(2)*y(-1)+(c(3)+c(4)*y(-1))/(1+exp(-c(5)*(y(-1))))

but, by the definition of the LSTAR model, c(5) must be positive. Is there any way to set restriction on c(5) before estimation ?

[EViews Glenn]

Reparameterize so that you use exp(c(5)) as your term in the model. You are then estimating the log of the term.