NONMEM Users Network Archive

Hosted by Cognigen

RE: [NMusers] bootstrap

From: Bill Denney <wdenney_at_humanpredictions.com>
Date: Fri, 10 May 2019 13:45:48 -0400

Hi Niurys,

The simplest method to ensure a good bootstrap is often to simplify the dat=
a
file by removing rows that should not be used for the modeling before
running the bootstrap. Notably, if you exclude an entire subject either
based on the ID column or another column, usually the boostrap will not wor=
k
correctly.

I believe that most if not all tools generate the bootstrap with new ID
column values (the ID is given a new sequential value based on sampling
order). If you exclude an entire subject based on another column, the you
will not have the expected number of subjects in the analysis because all
the bootstrap tools that I know of don't account for exclusions with making
the new data file.

If this doesn't help, giving more info will help. (What tool are you using
for bootstrap? What command line are you running? Can you share the model
and a snippet of the data?)

Thanks,

Bill

-----Original Message-----
From: owner-nmusers_at_globomaxnm.com <owner-nmusers_at_globomaxnm.com> On Behalf
Of Niurys.CS
Sent: Friday, May 10, 2019 12:17 PM
To: nmusers <nmusers_at_globomaxnm.com>
Subject: [NMusers] bootstrap

Dear nmusers,


I have a big doubt. When I used the bootstrap to evaluate my model, I had
some bugs. In my code I use IGNORE statements based on FLAGS for some
outliers. I don't know if I remove these IGNORE statements, the bootstrap
will run well. Can you give me some suggestions???????


Regards

Niurys de Castro Suárez

--

MSc Niurys de Castro Suárez
Profesor Asistente Farmacometría
Investigador Agregado
Departamento Farmacia
Instituto de Farmacia y Alimentos, Universidad de La Habana Cuba "Una
estrella brilla en la hora de nuestro encuentro"

Received on Fri May 10 2019 - 13:45:48 EDT

The NONMEM Users Network is maintained by ICON plc. Requests to subscribe to the network should be sent to: nmusers-request_at_iconplc.com. Once subscribed, you may contribute to the discussion by emailing: nmusers@globomaxnm.com.