In Simbiology, is the parameter estimates reliable, if I get the error " Iteration limit exceeded in the Laplacian algorithm. Returning result of the final iteration"

I use the nlme model with exponential error and FOCE method of likelihood estimation to fit the PK data. I got the highest loglikeihood with this setting. But I also got a error/warning that

"Iteration limit exceeded in the Laplacian algorithm. Returning result of the final iteration"

What should I do to overcome this error. Should I even try to overcome this.? With this error, if I try to report the parameter estimate is it authentic.?

I am also attaching the algorithm settings of iteration with this question for reference regarding my settings.

Answer 1

[Star Strider]

For that function, you can use the options created by the statset function to increase the number of iterations MaxIter beyond the default value of 200.

Try this:

opts = statset('nlmefit', 'MaxIter',1000);

to increase the iteration limit to 1000. (Assuming you are using the nlmefit function.)

NOTE — I have not tested this myself. It should work.

[Praveen Kumar M]

OK. I have tried to increase the iteration to 1000. I got this error now

Unable to find a step that will decrease SSE in Gauss-Newton algorithm., while solving PNLS. Returning results from last iteration.

I am using nlme inside SimBiology.

My one more part of question was whether the parameter estimated is valid even in the place of error/ warning. I am asking this as it is saying that "Returning result of the final iteration". So that means it should be valid right?

[Star Strider]

That likely means that the function has found what may be the global minimum. In any event, the function cannot further improve the parameter estimates.

Yes. It is valid. It is the best set of estimated parameters the function can calculate.

I do not have SimBiology, however I do have the other Toolboxes it apparently uses.

[Praveen Kumar M]

Thank so much Star Strider for swift and valuable answers.

[Star Strider]

As always, my pleasure!

Thank you for your compliment!

[Jeremy Huard]

Dear Praveen,

please avoid creating a new thread for consecutive questions coming from the same issue. I see that you already posted this one in another thread and that it resulted from trying the suggestion provided in this other thread.

Keeping successive questions in the same thread will help us understand the context and keep the original question in mind.

Thanks,

Jérémy

[Praveen Kumar M]

Dear Jeremy,

    Thanks for that info. I will not do it henceforth.

Regards.