Help with Bootstrapping Output

[Michael Kruger-2]

I am using the OMS (Output Management System) of SPSS to conduct
bootstrapped estimates of logistic regression coefficients and can use
some help in creating output files. To summarize the study, I have a
data set with some 220 pre-term neonates with 10 predictors that are
primarily ultrasound measures and some maternal demographics. I am using
a bootstrapping macro to generate estimated regression coefficients in a
lstepwise ogistic regression predicting whether the neonates will
survive or not. Because each bootstrap replication produces a different
model, I am saving the coefficeints into an SPSS data file and then
aggregating them to get my mean regression coefficients. My problem is
that I don't know how to include a replication counter variable that I
can use to select the last step coefficients from each replication. For
example, if I tell my bootstrap macro to run 100 reps, I need to
generate a variable for each set of replication coefficients (1 to 100)
and then be able to select the maximum step for each replication.


   Once I have accomplished this I can easily aggregate the selected
steps into mean values. So essentially I need some help in getting the
OMS system to generate a replication variable for all steps generated in
each replication and then some code for selecting the maximum step for
each replication. If you think you can help me out I can send you my OMS
macro that conducts the bootstrap. I created it from some sample code on
p. 160 of the 4th edition of the SPSS Programming and Data Management
Guide. Thanks in advance for your time and consideration.


Michael Kruger
Statistical Analyst
Dept. of OB/GYN
Wayne State University School of Medicine
Detroit, MI

=====================
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

[Marta Garcia-Granero]

Hi Michael

I don't know if it will help you, but adding the subcommand   /PRINT =
SUMMARY to the LOGISTIC REGRESSION command will force the output to
print the last step of the model only, avoiding thus the problem of
bootstrapping intermediate models. Anyway, how will you cope with the
problem of different variables being included in the different bootstrap
replications (it can happen with stepwise models)?. I would use
bootstrap for the final model, after having built it. I don't like
stepwise models, and they have been discussed (and very much criticized)
in this list in the past.

HTH,
Marta

> I am using the OMS (Output Management System) of SPSS to conduct
> bootstrapped estimates of logistic regression coefficients and can use
> some help in creating output files. To summarize the study, I have a
> data set with some 220 pre-term neonates with 10 predictors that are
> primarily ultrasound measures and some maternal demographics. I am using
> a bootstrapping macro to generate estimated regression coefficients in a
> lstepwise ogistic regression predicting whether the neonates will
> survive or not. Because each bootstrap replication produces a different
> model, I am saving the coefficeints into an SPSS data file and then
> aggregating them to get my mean regression coefficients. My problem is
> that I don't know how to include a replication counter variable that I
> can use to select the last step coefficients from each replication. For
> example, if I tell my bootstrap macro to run 100 reps, I need to
> generate a variable for each set of replication coefficients (1 to 100)
> and then be able to select the maximum step for each replication.
>
>
>   Once I have accomplished this I can easily aggregate the selected
> steps into mean values. So essentially I need some help in getting the
> OMS system to generate a replication variable for all steps generated in
> each replication and then some code for selecting the maximum step for
> each replication. If you think you can help me out I can send you my OMS
> macro that conducts the bootstrap. I created it from some sample code on
> p. 160 of the 4th edition of the SPSS Programming and Data Management
> Guide. Thanks in advance for your time and consideration.
>
>
> Michael Kruger
> Statistical Analyst
> Dept. of OB/GYN
> Wayne State University School of Medicine
> Detroit, MI
>
> =====================
> 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
>

--
For miscellaneous statistical stuff, visit:
http://gjyp.nl/marta/

=====================
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