Date: Fri, 31 Oct 1997 11:18:09 -0600
Reply-To: "Nichols, David" <nichols@SPSS.COM>
Sender: "SPSSX(r) Discussion" <SPSSX-L@UGA.CC.UGA.EDU>
From: "Nichols, David" <nichols@SPSS.COM>
The standard test would be testing the three way interaction
in a loglinear model, or the two way between ethnicity and
poor/non-poor in a logit model (which is the same thing).
David Nichols
Senior Support Statistician
SPSS Inc.
nichols@spss.com
>----------
>From: Robert Corwyn[SMTP:rbflynn@UALR.EDU]
>Sent: Friday, October 31, 1997 10:54 AM
>To: SPSSX-L@UGA.CC.UGA.EDU
>
>I apologize for posting this question a second time. I suspect the columns
>were off last time and it wasn't readable. I would greatly appreciate advise
>on the following...
>
>The following are odds obtained from a logit analysis with HOME items as the
>dependent variables. For each group, I ran a separate logit analysis with
>groups defined as, for example, white/non-poor=1 versus all others=0. Thus,
>Non-poor Whites were 2.02 times as likely to get credit for positive feelings
>as all other persons in the dataset.
>
>White Odds of receiving credit on a HOME item
>Non-Poor 2.02
>Poor .72
>
>Black
>Non-Poor 1.34
>Poor .48
>
>Hispanic
>Non-Poor 1.60
>Poor .55
>
>
>My question is this... While looking at the table, I noticed that the effect
>of poverty status appeared to be proportional across ethnic groups. For
>example,
>
>2.02 divided by .72= 2.806
>1.34 divided by .48= 2.792
>1.60 divided by .55= 2.909
>
>I have over 100 HOME items, and the proportional effect of poverty across
>ethnic groups is very consistent.
>
>Could someone suggest a statistical test to verify that the effects
>of poverty on my dependent variables is proportional across ethnic groups?
>I ran loglinear analyses testing for interactions between poverty status and
>ethnicity, and found no interaction effects (with Bonferonni adjustments).
>
>Thanks in advance
>Robert Flynn Corwyn
>
| 
