From: Martin Bergstrand <*martin.bergstrand*>

Date: Tue, 25 Aug 2009 16:04:49 +0200

Dear Nele,

Individual weighted residuals (IWRES) have nothing to do with the =

parameter

estimation in NONMEM. IWRES are calculated and put in table files =

purely as

a diagnostic variable. This way they can be helpful among other things =

to

design a suitable RUV model (and that is what I really think you need to

think about) .

Normally I would not want to add weighting to different parts of the =

data

based on my expectations. If I anticipated the case of lower residual

unexplained variability (RUV) with time I would include such a feature =

in my

RUV model and estimate parameters to describe it (and also evaluate =

it’s

significance). The easiest check to do is to estimate one magnitude of

residual error for early samples and one for late (see an example below =

for

a simple additive RUV) .

;------------------------------------------------------------------------=

---

---------------------------------------------

IF(TIME.LT.X) THEN ; X = Time limit dividing =

early

and late time points

W = THETA(x) ; THETA (x) = =

standard

deviation (SD) for RUV of early samples

ELSE

W = THETA(y) ; THETA (y) = SD for =

RUV of

late samples

ENDIF

Y = IPRED+W*EPS(1)

IRES = DV-IPRED

IWRES = IRES/W

$SIGMA 1 FIX ; Fixing SIGMA to =

variance

1 allows us to estimate the scaling factor W on standard deviation scale

;------------------------------------------------------------------------=

---

---------------------------------------------

I have no experience with a continuous RUV model that describes =

decreasing

RUV with time. If the data is informative enough I sure that it is =

possible

though. An example could look something like this:

W0 = THETA(x) ; SD =

for W

at time = 0

W1HL = THETA(y) ; =

Half-life of

time dependent RUV

WL = THETA(z) ; SD =

of

non time dependent RUV

W1K = LOG(2)/W1HL =

W1 = (W0- W2) * EXP(-W1K*TIME)

W = W1 + WL

Y = IPRED+W*EPS(1)

$SIGMA 1 FIX

;------------------------------------------------------------------------=

---

---------------------------------------------

I hope this is of some help to you.

Kind regards,

Martin Bergstrand, MSc, PhD student

-----------------------------------------------

Division of Pharmacokinetics and Drug Therapy,

Department of Pharmaceutical Biosciences, Uppsala University

-----------------------------------------------

P.O. Box 591

SE-751 24 Uppsala

Sweden

-----------------------------------------------

<mailto:martin.bergstrand

martin.bergstrand

-----------------------------------------------

Work: +46 18 471 4639

Mobile: +46 709 994 396

Fax: +46 18 471 4003

From: owner-nmusers

On

Behalf Of nele.kaessner

Sent: den 25 augusti 2009 14:21

To: nmusers

Subject: [NMusers] change calculation of WRES?

Dear nmusers,

I have a question which I hope is not too trivial for the group. I am

currently analyzing some data where I have more trust in values at late =

time

points. Therefore, I would like WRES to be a function of time, putting =

more

weight on late time points. In the help guide I found that WRES can be

influenced by SPTWO, however no real documentation exists (or I am not =

aware

of it) about how to use it. Does anybody have an example for me of how =

to

code this?

Moreover, I noticed that no matter how I describe my weighting for =

IWRES,

this does not at all seem to influence my objective function or =

parameter

values. When evaluating a model, most people consider anyway that WRES =

is

what counts, as IWRES are in most cases ok anyway. So my very simple

question is: If this does not influence any of this and I don't use =

IWRES to

decide if a model is good or bad, why bother at all to calculate them? I

noticed that the standard errors seem to change depending on which =

weighting

I use for IWRES. Can anybody explain this?

Thanks and best wishes

Nele

______________________________________________________________

Dr. Nele Käßner

Pharmacometrics -- Modeling and Simulation

Nycomed GmbH

Byk-Gulden-Str. 2

D-78467 Konstanz, Germany

Fon: (+49) 7531 / 84 - 4759

Fax: (+49) 7531 / 84 - 94759

mailto: nele.kaessner

http://www.nycomed.com

County Court: Freiburg, Commercial Register HRB 701257

Chairman Supervisory Board: Charles Depasse

Management Board: Dr. Barthold Piening, Gilbert Rademacher, Dr. Anders

Ullman

----------------------------------------------------------------------

Proprietary or confidential information belonging to Nycomed Group may

be contained in this message. If you are not the addressee indicated

in this message, please do not copy or deliver this message to anyone.

In such case, please destroy this message and notify the sender by

reply e-mail. Please advise the sender immediately if you or your

employer do not consent to Internet e-mail for messages of this kind.

Opinions, conclusions and other information in this message that

pertain to the sender's employer and its products and services

represent the opinion of the sender and do not necessarily represent

or reflect the views and opinions of the employer.

----------------------------------------------------------------------

Received on Tue Aug 25 2009 - 10:04:49 EDT

Date: Tue, 25 Aug 2009 16:04:49 +0200

Dear Nele,

Individual weighted residuals (IWRES) have nothing to do with the =

parameter

estimation in NONMEM. IWRES are calculated and put in table files =

purely as

a diagnostic variable. This way they can be helpful among other things =

to

design a suitable RUV model (and that is what I really think you need to

think about) .

Normally I would not want to add weighting to different parts of the =

data

based on my expectations. If I anticipated the case of lower residual

unexplained variability (RUV) with time I would include such a feature =

in my

RUV model and estimate parameters to describe it (and also evaluate =

it’s

significance). The easiest check to do is to estimate one magnitude of

residual error for early samples and one for late (see an example below =

for

a simple additive RUV) .

;------------------------------------------------------------------------=

---

---------------------------------------------

IF(TIME.LT.X) THEN ; X = Time limit dividing =

early

and late time points

W = THETA(x) ; THETA (x) = =

standard

deviation (SD) for RUV of early samples

ELSE

W = THETA(y) ; THETA (y) = SD for =

RUV of

late samples

ENDIF

Y = IPRED+W*EPS(1)

IRES = DV-IPRED

IWRES = IRES/W

$SIGMA 1 FIX ; Fixing SIGMA to =

variance

1 allows us to estimate the scaling factor W on standard deviation scale

;------------------------------------------------------------------------=

---

---------------------------------------------

I have no experience with a continuous RUV model that describes =

decreasing

RUV with time. If the data is informative enough I sure that it is =

possible

though. An example could look something like this:

W0 = THETA(x) ; SD =

for W

at time = 0

W1HL = THETA(y) ; =

Half-life of

time dependent RUV

WL = THETA(z) ; SD =

of

non time dependent RUV

W1K = LOG(2)/W1HL =

W1 = (W0- W2) * EXP(-W1K*TIME)

W = W1 + WL

Y = IPRED+W*EPS(1)

$SIGMA 1 FIX

;------------------------------------------------------------------------=

---

---------------------------------------------

I hope this is of some help to you.

Kind regards,

Martin Bergstrand, MSc, PhD student

-----------------------------------------------

Division of Pharmacokinetics and Drug Therapy,

Department of Pharmaceutical Biosciences, Uppsala University

-----------------------------------------------

P.O. Box 591

SE-751 24 Uppsala

Sweden

-----------------------------------------------

<mailto:martin.bergstrand

martin.bergstrand

-----------------------------------------------

Work: +46 18 471 4639

Mobile: +46 709 994 396

Fax: +46 18 471 4003

From: owner-nmusers

On

Behalf Of nele.kaessner

Sent: den 25 augusti 2009 14:21

To: nmusers

Subject: [NMusers] change calculation of WRES?

Dear nmusers,

I have a question which I hope is not too trivial for the group. I am

currently analyzing some data where I have more trust in values at late =

time

points. Therefore, I would like WRES to be a function of time, putting =

more

weight on late time points. In the help guide I found that WRES can be

influenced by SPTWO, however no real documentation exists (or I am not =

aware

of it) about how to use it. Does anybody have an example for me of how =

to

code this?

Moreover, I noticed that no matter how I describe my weighting for =

IWRES,

this does not at all seem to influence my objective function or =

parameter

values. When evaluating a model, most people consider anyway that WRES =

is

what counts, as IWRES are in most cases ok anyway. So my very simple

question is: If this does not influence any of this and I don't use =

IWRES to

decide if a model is good or bad, why bother at all to calculate them? I

noticed that the standard errors seem to change depending on which =

weighting

I use for IWRES. Can anybody explain this?

Thanks and best wishes

Nele

______________________________________________________________

Dr. Nele Käßner

Pharmacometrics -- Modeling and Simulation

Nycomed GmbH

Byk-Gulden-Str. 2

D-78467 Konstanz, Germany

Fon: (+49) 7531 / 84 - 4759

Fax: (+49) 7531 / 84 - 94759

mailto: nele.kaessner

http://www.nycomed.com

County Court: Freiburg, Commercial Register HRB 701257

Chairman Supervisory Board: Charles Depasse

Management Board: Dr. Barthold Piening, Gilbert Rademacher, Dr. Anders

Ullman

----------------------------------------------------------------------

Proprietary or confidential information belonging to Nycomed Group may

be contained in this message. If you are not the addressee indicated

in this message, please do not copy or deliver this message to anyone.

In such case, please destroy this message and notify the sender by

reply e-mail. Please advise the sender immediately if you or your

employer do not consent to Internet e-mail for messages of this kind.

Opinions, conclusions and other information in this message that

pertain to the sender's employer and its products and services

represent the opinion of the sender and do not necessarily represent

or reflect the views and opinions of the employer.

----------------------------------------------------------------------

Received on Tue Aug 25 2009 - 10:04:49 EDT