Date:         Wed, 28 Sep 2005 10:12:50 -0700
Reply-To:     Gregory Hildebrandt <greg.hildebrandt@sbcglobal.net>
Sender:       "SPSSX(r) Discussion" <SPSSX-L@LISTSERV.UGA.EDU>
From:         Gregory Hildebrandt <greg.hildebrandt@sbcglobal.net>
Subject:      Re: bootstrap
Comments: To: "Soeken, Karen L." <Soeken@son.umaryland.edu>
In-Reply-To:  <7FF0C67B2D894A4A8EA0DABFFC19F51403FC08E2@sonsvr04.son.umaryland.edu>
Content-Type: text/plain; charset=iso-8859-1

Karen:

SPSS "non-linear regression," which I believe is part of the regression  module, has an option for  computing the  standard  errors of  the regression coefficients using bootstrap.  One could  estimate  a linear regression  model  using the non-linear feature, and then do the test.

However, if  you want to design your own bootstrap procedure, then someone else will have to help with  the syntax program.

Greg

"Soeken, Karen L." <Soeken@son.umaryland.edu> wrote:
> Does anyone have experience using the syntax file for bootstrapping in multiple regression? I can do the bootstrapping when I use the sample syntax and sample datafile, but I run into difficulty with I try to use it with my own data file. The samples are drawn, but the estimates are not being saved into the temporary file. The syntax file notes that the regression coefficients will be save with _B and _Beta attached to the variable names, but the error message I get says that those new variable names (with _B and _Beta attached) are not recognized. They are recognized and the summary statistics are reported when I use the sample datafile, though.
>
> Karen L. Soeken, Ph.D.
> Professor, University of Maryland School of Nursing
> Tel: 410-706-7504
> e-mail: soeken@son.umaryland.edu
>