Problem with HESSIAN resolved

1 messages 1 people Latest: Jul 01, 2009

Problem with HESSIAN resolved

From: Jacob Brogren Date: July 01, 2009 technical
Hi all, The problem occured due to UVOL="." in the data set. NONMEM will then divide by zero. For some reason it worked with ADVAN4. I eventually used urine amounts instead of concentrations (and UVOL). Credit to Lena Friberg for the tip about UVOL being zero and to Leonid for a lot of extra help. /Jacob -------------------------------------------------------------------------- Confidentiality Notice: This message is private and may contain confidential and proprietary information. If you have received this message in error, please notify us and remove it from your system and note that you must not copy, distribute or take any action in reliance on it. Any unauthorized use or disclosure of the contents of this message is not permitted and may be unlawful.
Quoted reply history
-----Original Message----- From: [email protected] [mailto:[email protected]]on Behalf Of Brogren, Jacob Sent: 29 juni 2009 16:50 To: Leonid Gibiansky Cc: Nmusers (E-mail) Subject: RE: [NMusers] Problem with HESSIAN Hi Leonid, thank you for your input. I tried what you recommended. With FO NONMEM seems to enter a black hole: nothing gets out of the MONITORING OF SEARCH. I also removed ALAG but the problem is still there. Regarding ETAs, I have shuffled them around to see if it had an impact. But it didn't... This is a bit frustrating so I am realy grateful for all suggestions I can get from nmusers. Best Regards Jacob -------------------------------------------------------------------------- Confidentiality Notice: This message is private and may contain confidential and proprietary information. If you have received this message in error, please notify us and remove it from your system and note that you must not copy, distribute or take any action in reliance on it. Any unauthorized use or disclosure of the contents of this message is not permitted and may be unlawful. -----Original Message----- From: Leonid Gibiansky [mailto:[email protected]] Sent: 29 juni 2009 14:51 To: Brogren, Jacob Cc: Nmusers (E-mail) Subject: Re: [NMusers] Problem with HESSIAN Note also that your models are not identical: in ADVAN4 realization, you put random effect on ALAG1 while in ADVAN6 it is on F1. Try to remove ALAG1 (or fix it at 0.0001 or some small value, before the first observation). The error could be related to ALAG1 difficulties. Also, start ADVAN6 problem with the final values of ADVAN4 problem (could be with MAXEVAL=0), just to get something converged. It could be even FO: just to start with something. Then, after you see that the bulk part is correct, you can move on to sort out details. Leonid -------------------------------------- Leonid Gibiansky, Ph.D. President, QuantPharm LLC web: www.quantpharm.com e-mail: LGibiansky at quantpharm.com tel: (301) 767 5566 Brogren, Jacob wrote: > Hi Leonid, > > I tried to change to ADVAN8/ADVAN9 and TOL=8. I get the same error message. > > By the way, Martin Brgstrand tiped me about (F+DEL)**2 instead of F**2+DEL > since NONMEM might not like the concept of 0 raised to the power of 2. > However, this didn't change the outcome. > > Best Regards > > Jacob > > > -------------------------------------------------------------------------- > Confidentiality Notice: This message is private and may contain confidential > and proprietary information. If you have received this message in error, > please notify us and remove it from your system and note that you must not > copy, distribute or take any action in reliance on it. Any unauthorized use > or disclosure of the contents of this message is not permitted and may be > unlawful. > > -----Original Message----- > From: Leonid Gibiansky [mailto:[email protected]] > Sent: 29 juni 2009 13:31 > To: Brogren, Jacob > Cc: Nmusers (E-mail) > Subject: Re: [NMusers] Problem with HESSIAN > > > Jacob, > DEL part is not needed (unless you do log-transform; and even then, > there are better methods). > Try TOL=8 and/or ADVAN8 ADVAN9 > > Leonid > > -------------------------------------- > Leonid Gibiansky, Ph.D. > President, QuantPharm LLC > web: www.quantpharm.com > e-mail: LGibiansky at quantpharm.com > tel: (301) 767 5566 > > > > > Brogren, Jacob wrote: >> Hi, thank's for the advice. Unfortunately neither "DEL=0.01" nor "S3=V3" >> resolved the issue. I've also tried log-transforming both sides with no >> improvement. >> >> Regards >> >> Jacob >> >> >