Threshold determination

[hsmellin]

Hi,

i am looking to estimate a TAR model and to search for the threshold value of the dependent variable using the Chan (1993) methodology*.

I try the following code:


vector(67) ssr_result

for !i=1 to 67

genr h_!i=0

genr tau_!i=@elem(y, !i)

smpl 2 67 if y(-1)>=tau_!i
genr h_!i=1

smpl 2 67
equation tar_{!i}.ls d(y)=c(1)*h_!i*((y(-1)-tau_!i))+c(2)*(1-h_!i)*((y(-1)-tau_!i))+c(3)*d(y(-1))

scalar ssr_{!i}=@ssr
ssr_result(!i)=@ssr


next

(I have sorted and trimmed the observations before running this code). However, i get an error message for "genr tau ...." i think there may be more errors with this code; i am very new to programming so apologies if the whole code is incorrect!

*Chan, K. S. (1993), "Consistency and Limiting Distribution of the Least Squares Estimator of a Threshold Autoregressive Model," The Annals of Statistics, 21, 520-533

[ninjagirl]

Hello,

Have you found the answer to your question? I am looking for that as well. Do help me out if you have found a solution.


I am trying to use the TARCOINT add-in

Regards,
Prarthana

[ewa]

Hello,

The add-in is very helpful. I managed to get all the estimates I wanted. Just wanted to make sure, the output does not provide information criteria results. Is there any way to get them to compare models?

Best regards,

Ewa

[trubador]

Lag-length selection already uses information criteria to decide the appropriate lag. I am not sure, but you may estimate the final model in light of the add-in's output and then compute/extract the information criteria ex-post for model comparison purposes.