Hessian error

For problems related to Pmetrics execution, not for modeling or dose optimization issues.
Post Reply
SONIALU
Posts: 13
Joined: Fri Mar 04, 2016 4:26 am

Hessian error

Post by SONIALU » Fri Dec 07, 2018 9:20 am

Hello!

I am trying to model my data with more than 300 hundred patients.
Despite the predictions are good, the Hessian error continuously appears.
I have been trying to take out the patients who suggested that they could be causing the error (one each in one run) but the Hessian error still appears.
I have also tried to increase gamma.

Any other idea?

Thank you very much
Sonia

User avatar
mneely
Site Admin
Posts: 234
Joined: Tue Nov 08, 2011 11:08 am
Contact:

Re: Hessian error

Post by mneely » Tue Dec 18, 2018 11:01 am

A Hessian error means that the model is not serving some or all patients well, among other things. Have you identified problem patients that, when removed, allow the run to proceed?

SONIALU
Posts: 13
Joined: Fri Mar 04, 2016 4:26 am

Re: Hessian error

Post by SONIALU » Sat Dec 29, 2018 7:58 am

Dear Michael,

yes. I have been trying to exclude all those patients but in the last run Pmetrics is telling me
that the patients that is causing the error in number 304 and this patient is not included in the final run.ç
So, I do not know what else can I do.

Thanks
Sonia

User avatar
mneely
Site Admin
Posts: 234
Joined: Tue Nov 08, 2011 11:08 am
Contact:

Re: Hessian error

Post by mneely » Sat Dec 29, 2018 10:05 am

If you remove patient 304 does the run complete?

SONIALU
Posts: 13
Joined: Fri Mar 04, 2016 4:26 am

Re: Hessian error

Post by SONIALU » Tue Jan 08, 2019 9:49 am

It is strange... because the patient 304 had been already excluded from the previous run.. before Pmetrics gave me this message.
Sonia

User avatar
mneely
Site Admin
Posts: 234
Joined: Tue Nov 08, 2011 11:08 am
Contact:

Re: Hessian error

Post by mneely » Tue Jan 08, 2019 10:26 am

Pmetrics numbers the patients sequentially. You have to look at the ID numbers that correspond with the subject number. For example, subject 304 could have an ID of AX440. If you removed the 304th subject, then the subject who was previously the 305th is now the 304th.

SONIALU
Posts: 13
Joined: Fri Mar 04, 2016 4:26 am

Re: Hessian error

Post by SONIALU » Mon Jan 14, 2019 3:14 am

Dear Michael,
Thanks a lot,

then... after removing some patients, the ID does not correspond with the number that Pmetrics is referring?
Sonia

I have also another questions, when Pmetrics is doing NPDE or Simulations, is it using a boostrap method with replacement?

Post Reply