Date: Tue, 8 Nov 2005 18:17:15 +0100 Marta García-Granero "SPSSX(r) Discussion" Marta García-Granero Asesoría Bioestadística Re: Basic nonparametric setup q text/plain; charset=ISO-8859-1

Hello Fred

FERPD> I don¹t use nonparametrics very often and am having difficulty with what I FERPD> assume is a basic problem.

FERPD> Scenario: Study of helping behavior under 2 crossed conditions: Gender FERPD> and Risk. Data are setup like so:

FERPD> Gender Risk Help?

FERPD> M High No FERPD> M High No FERPD> M Low Yes FERPD> M Low Yes FERPD> F High Yes FERPD> F High Yes FERPD> F Low Yes FERPD> F Low Yes

FERPD> First, is this the proper way to setup the data?

Yes

FERPD> Second, what is the proper FERPD> nonparametric test for this scenario? I am most interested in an FERPD> interaction between Gender and Risk, if that is even possible. Otherwise, FERPD> whatever will help identify main effects of gender and risk on helping FERPD> behavior.

Use multiple logistic regression or a stratified crosstabs:

* Simulated data (replace by your own) *. DATA LIST LIST/Gender(A1) Risk(A4) Help(A3). BEGIN DATA M High No M High Yes M Low Yes M Low Yes F High No F High Yes F Low Yes F Low Yes M High Yes M High Yes M Low No M Low No F High No F High No F Low No F Low No END DATA.

* Using logistic regression *. LOGISTIC REGRESSION VARIABLES=Help /METHOD = ENTER Gender Risk /METHOD = ENTER Gender*Risk /CONTRAST (Gender)=Indicator /CONTRAST (Risk)=Indicator /PRINT = CI(95).

* Using a stratified analysis *. CROSSTABS /TABLES=Help BY Risk BY Gender /FORMAT= AVALUE TABLES /STATISTIC=CHISQ RISK CMH(1) /CELLS= COUNT /COUNT ROUND CELL .

Both methods will test the significance of the interaction between gender and risk

Regards, Marta García-Granero mailto:biostatistics@terra.es

Back to: Top of message | Previous page | Main SPSSX-L page