# RE: Two dependent variables

From: Serge Guzy <GUZY>
Date: Wed, 31 Jan 2007 11:20:28 -0800

If we assume that the residual variances (sigma) across the different
responses are independent, then the computation of the objective
function is similar than the one obtained with one response. There is no
need to change what you call the default objective function.
Serge Guzy

-----Original Message-----
From: owner-nmusers
On Behalf Of Zheng Lu
Sent: Wednesday, January 31, 2007 9:11 AM
To: Nick Holford
Cc: nmusers
Subject: Re: [NMusers] Two dependent variables

Dear all:

As to this kind of situation(at least two types of observations or same
type of observations from at least two different assay), DO we need
change our default objective function?

Because I am thinking we will estimate our parameters based on
multivariate normal distribution(at least bivariate normal
distribution) instead of univariate normal distribution(only one type of
observation), so we will have different variance across the observations
and we will have covariance issue which is not zero.
thanks a lot.

Zheng Lu
Department of pharmaceutical sciences
College of Pharmacy
University of Michigan
428 Church Street,
Ann Arbor, MI 48109-1065
Tel: 734-763-7608(o), 734-647-1427(o), 734-647-1437(L) Fax:
734-615-6162(o), 734-763-3438(o)

Quoting Nick Holford <n.holford

> Hi,
>
> The 'FLAG' method is a trick that only works with 2 DVs. It is also a
> bit inefficient because it requires that both the PK and PD prediction

> are calculated but only one is used.
>
> A more general method uses IF THEN blocks e.g. if you have a data item

> DVID with the value 1 with a conc observation and the value 2 with an
> effect observation of one type and value of 3 with an effect
> observation of a different type:
>
> CONC=F
> IF (DVID.EQ.1) THEN
> Y=CONC + ERR(1)
> ENDIF
> IF (DVID.EQ.2) THEN
> Y=EMAX2*CONC/(EC502+CONC) + ERR(2)
> ENDIF
> IF (DVID.EQ.3) THEN
> Y=EMAX3*CONC/(EC503+CONC) + ERR(2)
> ENDIF
>
> Nick
>
>>
>> Yes, you could do a simultaneous fitting with the same control

>> stream in NM. For example you could
>> have a dummy variable (Flag) that tells NM which likelihood
>> function goes with which data. See below
>>
>>
>>
>> Flag 0 = PK, Flag 1=PD
>>
>>
>>
>> \$ERROR
>>
>> Y=CONC*(1-Flag) + EFF* Flag
>>
>>
>>
>> Hopefully this would help.
>>
>>
>>
>> Best,
>>
>> Nidal AL-Huniti, PhD
>>
>>
>> On 1/30/07, navin goyal <navin1180
>>
>> Hello,
>> I was wondering if I could model two DVs simultaneously
>> with the same control stream.
>> Like have to equations A=(equation)
>> B= (equation)
>>
>> I have data for two different drugs from same patients
>> Could I do it nomem ? and if yes how do I do it?
>>
>> Any help is appreciated,
>> Thanks
>>
>>
>>
>>
>>
>
>
>
> --
> Nick Holford, Dept Pharmacology & Clinical Pharmacology University of
> Auckland, 85 Park Rd, Private Bag 92019, Auckland, New Zealand
> email:n.holford
> http://www.health.auckland.ac.nz/pharmacology/staff/nholford/
>
>
>

--
The information contained in this email message may contain confidential or=
legally privileged information and is intended solely for the use of the n=
amed recipient(s). No confidentiality or privilege is waived or lost by an=
y transmission error. If the reader of this message is not the intended rec=
ipient, please immediately delete the e-mail and all copies of it from your=
system, destroy any hard copies of it and notify the sender either by tele=
phone or return e-mail. Any direct or indirect use, disclosure, distributi=
on, printing, or copying of any part of this message is prohibited. Any vi=
ews expressed in this message are those of the individual sender, except wh=
ere the message states otherwise and the sender is authorized to state them=
to be the views of XOMA.
Received on Wed Jan 31 2007 - 14:20:28 EST

The NONMEM Users Network is maintained by ICON plc. Requests to subscribe to the network should be sent to: nmusers-request@iconplc.com.

Once subscribed, you may contribute to the discussion by emailing: nmusers@globomaxnm.com.