EViews.com

Hi,

I am currently working on TAR and M-TAR model. I have following issues, it would be really helpful if someone could guide me on that:
1. Could anyone please guide me how to get the exact value of AIC,SIC and RMSE for the model generated by the tarcoint package of EViews.
2. When I use more than 2 variables in the endogenous list (and also when as a group) the add in shows an error saying Vtemp not defined. There is no such thing mentioned in the help file of tarcoint. Could anyone please tell me why the problem is occurring and how it can be solved.
3. After estimating the asymmetric equation of the error term how does one go about estimating the actual error correction system (see equation 20. 21) in the original Enders Silos Paper https://virtualtour.wlu.ca/documents/14 ... s_jbes.pdf

Thanks in advance.

1) AIC and SIC are only used during selection of appropriate lag length. Residual sum of squares (RSS) are used during the search for threshold value. The package does not store any of these values, but since you'll have the final model you may be able to compute them ex-post.

2) The add-in works fine with more than 2 variables: 3) The same logic applies here as well. Add-in estimates the threshold values along with the appropriate lag length. So you should be able to estimate the equations you cite after running the add-in.

Thanks trubador for your reply.

But the machine is giving me the same error again and again when I am entering more than 2 variables in the exogenous list and also when I am entering the other exogenous variables as a group. The error says "vtemp not defined."

Could you please post your workfile?

PFA, the workfile.

You are running out of degrees-of-freedom. You can either change the number of lags to be tested something smaller than the default value of 10 or adjust the trimming ratio something smaller than the default value of 0.15.

Thank you trubador. That was indeed a very silly mistake that I was making. I believe that probably the error message wasn't descriptive enough and hence I could not identify the issue. I would like if the error message could be redefined to avoid such naive mistakes by anyone like me in future, or mention them in the help file.

I also wanted to ask if it is possible to generate the granger coefficient of coherence based on cross spectral correlation using some add-in or an already written code available for the same.

Looking forward to your reply.

1) As you can realize, more than one mistake may lead to same type of error ("singular matrix" for instance) and it is not efficient/practical to issue a separate warning message for each case. The main purpose of add-ins (at least the ones I have written) is to extend the current functionality of EViews while providing the users with similar (but unfortunately not identical) ease of use. Since I am simply a user like yourself not a software developer, you cannot expect me to produce fully fledged add-ins.

2) I do not have (or know of) any codes that compute the coefficient you refer to.

Thanks for your help and I am really sorry that I mistook you for a software developer working for EViews. Seeing your level of activeness and knowledge with regard to EViews on this platform, I guess, any new entrant to the forum is bound to make such a mistake. I really apologize for my mistake.

Once again thanks for your help.

No apologies are necessary. I was just making sure that all mistakes and deficiencies are mine and should not be associated with EViews' developer team. Sort of expectation management :)