Help to estimate non-linear asymmetric mean reversion test
Posted: Fri Apr 13, 2007 11:19 amDear sir,
I need a desparate help. I have heard you people help a lot to RATS users. I am using RATS(v.6).Without this estimation I cant proceed with my PhD thesis.
I would like to test asymmetric non-linear mean reversion documented by--
"Tests for symmetric and asymmetric nonlinear mean reversion"-- Sollis, Leybourne and Newbold.
Journal of money credit and banking, vol-34, no.3, 2002.
zt=yt-mean
where
yt is the price series
∆ zt= α st(γ1, γ2, zt-1)*zt-1 + ∑ ßi* ∆ zt-i + et
It= 1 if zt-1 is positive other wise zero.
st(γ1, γ2, zt-1) = 1/[1+exp{-(γ1**2)*((zt-1)**2)*It-(γ2**2)*((zt-1)**2)*(1-It)}] - 0.5
Note mean need to be estimated within the non-linear estimate process (not seperately).
I did coding but its not working. It gives strange estimate.
Please help me out.
Thanking you
saikat sarkar
PhD student
Tampere University,
Tampere, Finland
I need a desparate help. I have heard you people help a lot to RATS users. I am using RATS(v.6).Without this estimation I cant proceed with my PhD thesis.
I would like to test asymmetric non-linear mean reversion documented by--
"Tests for symmetric and asymmetric nonlinear mean reversion"-- Sollis, Leybourne and Newbold.
Journal of money credit and banking, vol-34, no.3, 2002.
zt=yt-mean
where
yt is the price series
∆ zt= α st(γ1, γ2, zt-1)*zt-1 + ∑ ßi* ∆ zt-i + et
It= 1 if zt-1 is positive other wise zero.
st(γ1, γ2, zt-1) = 1/[1+exp{-(γ1**2)*((zt-1)**2)*It-(γ2**2)*((zt-1)**2)*(1-It)}] - 0.5
Note mean need to be estimated within the non-linear estimate process (not seperately).
I did coding but its not working. It gives strange estimate.
Please help me out.
Thanking you
saikat sarkar
PhD student
Tampere University,
Tampere, Finland