Dear NMUsers
I have a problem and hope someone might be able to help. In summary, is it
possible for NONMEM to model an OMEGA matrix with four elements (W,X,Y,Z)
such that two of the covariances are fixed to zero and the rest freely
determined? I realise NONMEM can model BAND matrices, but I can't see how to
fix just two covariances to zero. The final matrix structure I'm looking for
would look something like.
Sw
C(w,x) Sx
0 C(x,y) Sy
0 C(x,z) C(y,z) Sz
I hope that makes sense. If so, is it possible?
Many thanks
Gavin
__________________________________________________
Dr Gavin E Jarvis MA(Cantab) MA PhD VetMB MRCVS
University Lecturer in Veterinary Anatomy
Department of Physiology, Development & Neuroscience
Physiological Laboratory
Downing Street
Cambridge
CB2 3EG
Tel: +44 (0) 1223 333745
Fellow and College Lecturer in Pharmacology
Tutor for Graduate Students
Selwyn College
Cambridge
CB3 9DQ
Tel: +44 (0) 1223 761303
Email: <mailto:[email protected]> [email protected]
Web: http://www.pdn.cam.ac.uk/staff/jarvis www.pdn.cam.ac.uk/staff/jarvis
Twitter: @GavinEJarvis
Zeros in OMEGA matrix
3 messages
3 people
Latest: Jul 03, 2015
Gavin,
NM can not estimate the type of matrix that you have described.
I would suggest adding C(w,y). It may be poorly estimated but you can always comment that it was added to allow estimation of the others.
Hope it works out for you,
Luann
Quoted reply history
On 7/3/2015 9:58 AM, Gavin Jarvis wrote:
> Dear NMUsers
>
> I have a problem and hope someone might be able to help. In summary, is it possible for NONMEM to model an OMEGA matrix with four elements (W,X,Y,Z) such that two of the covariances are fixed to zero and the rest freely determined? I realise NONMEM can model BAND matrices, but I can't see how to fix just two covariances to zero. The final matrix structure I'm looking for would look something like...
>
> Sw
>
> C(w,x) Sx
>
> 0 C(x,y) Sy
>
> 0 C(x,z) C(y,z) Sz
>
> I hope that makes sense. If so, is it possible?
>
> Many thanks
>
> Gavin
>
> __________________________________________________
>
> *Dr Gavin E Jarvis MA**(Cantab)**MA PhD VetMB MRCVS*
>
> University Lecturer in Veterinary Anatomy
>
> Department of Physiology, Development & Neuroscience
>
> Physiological Laboratory
>
> Downing Street
>
> Cambridge
>
> CB2 3EG
>
> Tel: +44 (0) 1223 333745
>
> Fellow and College Lecturer in Pharmacology
>
> Tutor for Graduate Students
>
> Selwyn College
>
> Cambridge
>
> CB3 9DQ
>
> Tel: +44 (0) 1223 761303
>
> Email: [email protected] <mailto:[email protected]>
>
> Web: www.pdn.cam.ac.uk/staff/jarvis < http://www.pdn.cam.ac.uk/staff/jarvis >
>
> Twitter: @GavinEJarvis
--
*Luann Phillips*| Director, Pharmacometrics
Cognigen Corporation, a wholly owned subsidiary of Simulations Plus, Inc.
1780 Wehrle Drive, Suite 110 | Buffalo, NY 14221-7000
Phone: 716.633.3463 x236 | Fax: 716.633.7404 | [email protected] < mailto: [email protected] >
You may define
MYETA1 = C11*ETA(1)
MYETA2 = C12*ETA(1)+C22*ETA(2)
MYETA3 = C23*ETA(2)+C33*ETA(3)
MYETA4 = C24*ETA(2)+C34*ETA(3)+C44*ETA(4)
with diagonal (and fixed to unit matrix) OMEGA matrix of ETA(1:4)
Then OMEGA for MYETA[1:4] matrix is
C11*C11
C11*C12 C12*C12+C22*C22
0 C22*C23 C23*C23+C33*C33
0 C22*C24 C23*C24+C33*C34 C24*C24+C34*C34+C44*C44
I am not sure whether it makes sense to go that far to get an extra 0 or extra flexibility (relative to the band matrices with 1 or 3 zeros), but technically, this is equivalent to the matrix that you need.
Another (simpler) option is to define
MYETA1 = ETA(1)
MYETA2 = ETA(2)
MYETA3 = ETA(3)
MYETA4 = ETA(4)+C24*ETA(2)
with the band matrix for ETAs:
Sw
C(w,x) Sx
0 C(x,y) Sy
0 0 C(y,z) Sz
The term C24*ETA(2) will add extra correlation (2-4) for MYETA matrix that is forbidden by the band matrix for ETA(1:4).
Leonid
--------------------------------------
Leonid Gibiansky, Ph.D.
President, QuantPharm LLC
web: www.quantpharm.com
e-mail: LGibiansky at quantpharm.com
tel: (301) 767 5566
Quoted reply history
On 7/3/2015 9:58 AM, Gavin Jarvis wrote:
> Dear NMUsers
>
> I have a problem and hope someone might be able to help. In summary, is
> it possible for NONMEM to model an OMEGA matrix with four elements
> (W,X,Y,Z) such that two of the covariances are fixed to zero and the
> rest freely determined? I realise NONMEM can model BAND matrices, but I
> can’t see how to fix just two covariances to zero. The final matrix
> structure I’m looking for would look something like…
>
> Sw
>
> C(w,x) Sx
>
> 0 C(x,y) Sy
>
> 0 C(x,z) C(y,z) Sz
>
> I hope that makes sense. If so, is it possible?
>
> Many thanks
>
> Gavin
>
> __________________________________________________
>
> *Dr Gavin E Jarvis MA**(Cantab)**MA PhD VetMB MRCVS*
>
> University Lecturer in Veterinary Anatomy
>
> Department of Physiology, Development & Neuroscience
>
> Physiological Laboratory
>
> Downing Street
>
> Cambridge
>
> CB2 3EG
>
> Tel: +44 (0) 1223 333745
>
> Fellow and College Lecturer in Pharmacology
>
> Tutor for Graduate Students
>
> Selwyn College
>
> Cambridge
>
> CB3 9DQ
>
> Tel: +44 (0) 1223 761303
>
> Email: [email protected] <mailto:[email protected]>
>
> Web: www.pdn.cam.ac.uk/staff/jarvis http://www.pdn.cam.ac.uk/staff/jarvis
>
> Twitter: @GavinEJarvis