estimating LSTR using GMM and NLS
Posted: Wed Jul 27, 2011 1:43 pm
Hi,
I have a simple LSTR (logistic smooth transition) model which I know can be estimated using GMM where moments are set to replicate NLS.
Basically, I have
y= [1-f(regime(-1))]*beta1'*x +(f(regime(-1))*beta2'*x
where
f(regime(-1)) = 1/1+exp (-gamma'(regime(-1)-c)
and
x={a,b, y(-1), constant}.
x has beta1 if regime variable "regime"<0.3 and beta2 if regime>0.3.
Extra info:
gamma = 2.5
bandwidth = 2
Despite knowing which method to use and looking through the manual, I am still confused about how to approach this problem. Any ideas on what should I put in what box?
I have access to Eviews 6 and 7.
Thanks!
M.
I have a simple LSTR (logistic smooth transition) model which I know can be estimated using GMM where moments are set to replicate NLS.
Basically, I have
y= [1-f(regime(-1))]*beta1'*x +(f(regime(-1))*beta2'*x
where
f(regime(-1)) = 1/1+exp (-gamma'(regime(-1)-c)
and
x={a,b, y(-1), constant}.
x has beta1 if regime variable "regime"<0.3 and beta2 if regime>0.3.
Extra info:
gamma = 2.5
bandwidth = 2
Despite knowing which method to use and looking through the manual, I am still confused about how to approach this problem. Any ideas on what should I put in what box?
I have access to Eviews 6 and 7.
Thanks!
M.