Problem with LSTAR model estimation
Posted: Fri Jul 09, 2010 2:33 am
Hello,
I'm estimating a logistic STAR of order 3 for a time series with the following equation object:
equation LSTAR3.ls u_tr = c(1) + c(2) * u_tr(-1) + c(3) * u_tr(-2) + c(4) * u_tr(-3) + (1/(1+exp( - (c(5)^2) * (s - c(6)) ) ) ) * ( c(7) + c(8) * u_tr(-1) + c(9) * u_tr(-2) + c(10) * u_tr(-3) )
(s is my transition variable)
However, I meet two problems:
1. The output gives me the message "WARNING: Singular covariance - coefficients are not unique". Moreover "Std. Error", "t-Statistic" and "Prob." are not available. How can I do to avoid this problem?
2. The specification will become very large with 5-6 lags, is there a shorter way to write the equation?
Hoping someone could help me.
Regards,
Richard
I'm estimating a logistic STAR of order 3 for a time series with the following equation object:
equation LSTAR3.ls u_tr = c(1) + c(2) * u_tr(-1) + c(3) * u_tr(-2) + c(4) * u_tr(-3) + (1/(1+exp( - (c(5)^2) * (s - c(6)) ) ) ) * ( c(7) + c(8) * u_tr(-1) + c(9) * u_tr(-2) + c(10) * u_tr(-3) )
(s is my transition variable)
However, I meet two problems:
1. The output gives me the message "WARNING: Singular covariance - coefficients are not unique". Moreover "Std. Error", "t-Statistic" and "Prob." are not available. How can I do to avoid this problem?
2. The specification will become very large with 5-6 lags, is there a shorter way to write the equation?
Hoping someone could help me.
Regards,
Richard