The RATS Software Forum

I have got a question to ask you. I was trying to use RATS to do some nonlinear estimations. During my grid search process, I need to specify different initial values for the parameters. As you know, some of the initial values may work and some may not. However, RATS will stop executing the loops and report an error message whenever a mistake is detected (say, a non-PSD. covariance matrix occurs due to bad initial values).

Whenever this happens, I will have to restart the program manually. This is very time-consuming. I am wondering if there is a command in RATS, as in MATLAB or EViews, which controls the maximum number of errors that RATS allows in the program. Before this maximum number is reached, RATS will not terminate the program.

Can I do this in RATS?

If you have an error which prevents an instruction from executing, and it's inside a loop or procedure, then RATS will terminate the loop since it doesn't necessarily know whether it even knows what to do next if an instruction fails. Is the condition which causes failure something for which you could test?

Hi Tom, I actually have a similar question as posted above.

What I did was simply using different initial values to estimate a nonlinear system. Some of the initial values worked but most failed due to various reasons. I wrote a loop so that RATS could try different initial values and repeat the same estimation procedure each time. Whenever an error message occurs, RATS stops as you said. Could I ask RATS to tolerate these errors for fixed number of time before it stop executing other attempts of initial values? Many thanks.