Hey all!
I am working on my master thesis and I am desperate with my model.
It looks as following:
Y(t) = ?1*X1(t) + ?2*X2(t) + ?*(?1*((1+c)/(?+c))+?2)*IE(t) -
?2*?*((1+c)/(?+c))*(?+g)* IE(t-1)
note: c and g is a constant value
The problem I encounter is that between IE(t) and IE(t-1) there is strong
linear correlation (autocorrelation). How can I solve this problem? Of
utterly importance is to have finally a significant coefficient ? and ?
which is than used for a consecutive model. However, I get either no
significant values for ? and ?, or for one of the two some unrealistic
values.
Is there an option to combine both in using some non-linear time lagged
model, time series or plugged in autoregression?
A following up question would be how to place penalties for this model. I
would like to restrict values for ? and ? between 0 and 0.5 and add
penalties when they come closer to the boundaries.
I really need some help. Because I am stuck with it for the last two weeks
and don't know how to go about it.
Thanks for the support
Cheers,
Bob
--
View this message in context:
http://r.789695.n4.nabble.com/Autocorrelation-in-non-linear-regression-model-tp3385647p3385647.html
Sent from the R help mailing list archive at Nabble.com.
