Monte Carlo simulation and Logistic regression

[Michael Kruger]

I was wondering if anyone on the list has used SPSS to conduct Monte
Carlo simulation with logistic regression. I have a physician who wants
to use about six predictors; ultrasound measures, gestational age,
ethnicicity and such to predict death of preterm neonates. We have about
300 neonates that meet a specific criteria in our dataset, of which
perhaps 100 have the event (death) and 200 who don't have the event (no
death). and he wants to use Monte Carlo resampling methods to get a
better estimate of the logistic regression parameters; betas and ORs.
The idea is that the measures will be more robust (precise or smaller
SEs) by resampling the 300 cases and provide a better model. We are
talking about something on the order of 1000 resamplings of the dataset.


I would appreciate any comments on this idea and how best to go about it
using SPSS. Thanks.

Michael Kruger
Statistical Analyst
Wayne State University School of Medicine
C.S. Mott Center
Dept. of OB/GYN

=====================
To manage your subscription to SPSSX-L, send a message to
[hidden email] (not to SPSSX-L), with no body text except the
command. To leave the list, send the command
SIGNOFF SPSSX-L
For a list of commands to manage subscriptions, send the command
INFO REFCARD

[SR Millis-3]

I prefer to perform model validation using the boostrap for obtaining nearly unbiased estimates of future model performance. Bootstrapping validates the process that was used to fit the original model. It provides an estimate of the expected value of the optimism , which when subtracted from the original index, provides an estimate of the expected bias-corrected estimates.

This is easily implemented in S-Plus/R using Harrell's Design library.


Scott R Millis, PhD, MEd, ABPP (CN,CL,RP), CStat
Professor & Director of Research
Dept of Physical Medicine & Rehabilitation
Wayne State University School of Medicine
261 Mack Blvd
Detroit, MI 48201
Email:  [hidden email]
Tel: [phone]
Fax: [phone]


--- On Thu, 7/10/08, Michael Kruger <[hidden email]> wrote:

> From: Michael Kruger <[hidden email]>
> Subject: Monte Carlo simulation and Logistic regression
> To: [hidden email]
> Date: Thursday, July 10, 2008, 6:41 PM
> I was wondering if anyone on the list has used SPSS to
> conduct Monte
> Carlo simulation with logistic regression. I have a
> physician who wants
> to use about six predictors; ultrasound measures,
> gestational age,
> ethnicicity and such to predict death of preterm neonates.
> We have about
> 300 neonates that meet a specific criteria in our dataset,
> of which
> perhaps 100 have the event (death) and 200 who don't
> have the event (no
> death). and he wants to use Monte Carlo resampling methods
> to get a
> better estimate of the logistic regression parameters;
> betas and ORs.
> The idea is that the measures will be more robust (precise
> or smaller
> SEs) by resampling the 300 cases and provide a better
> model. We are
> talking about something on the order of 1000 resamplings of
> the dataset.
>
>
> I would appreciate any comments on this idea and how best
> to go about it
> using SPSS. Thanks.
>
> Michael Kruger
> Statistical Analyst
> Wayne State University School of Medicine
> C.S. Mott Center
> Dept. of OB/GYN
>
> =====================
> To manage your subscription to SPSSX-L, send a message to
> [hidden email] (not to SPSSX-L), with no body
> text except the
> command. To leave the list, send the command
> SIGNOFF SPSSX-L
> For a list of commands to manage subscriptions, send the
> command
> INFO REFCARD

=====================
To manage your subscription to SPSSX-L, send a message to
[hidden email] (not to SPSSX-L), with no body text except the
command. To leave the list, send the command
SIGNOFF SPSSX-L
For a list of commands to manage subscriptions, send the command
INFO REFCARD