Date: Thu, 17 Nov 2005 15:53:43 -0500
Reply-To: "Woodring, Jonathan" <JWoodring@PPV.org>
Sender: "SPSSX(r) Discussion" <SPSSX-L@LISTSERV.UGA.EDU>
From: "Woodring, Jonathan" <JWoodring@PPV.org>
Subject: Re: Analysis of categorical data - is Chi-Square my only option?
Content-Type: text/plain; charset="us-ascii"
Logistic regression is the "kitchen sink" approach you seek, though you
should certainly educate yourself on the assumptions of the model and
appropriate diagnostics. If your "quit" variable is simply a binary
variable without taking time into account (i.e. someone quit or not over
a given period of time), then logistic regression will supply you with
odds ratios for each of your predictors (you should also read up on how
to interpret odds ratios, as they are often misinterpreted). But if you
have data on WHEN someone quits, or can create such a variable given
your data set, then survival analysis may be a possibility.
-----Original Message-----
From: SPSSX(r) Discussion [mailto:SPSSX-L@LISTSERV.UGA.EDU] On Behalf Of
E. Griffin
Sent: Thursday, November 17, 2005 3:33 PM
To: SPSSX-L@LISTSERV.UGA.EDU
Subject: Re: Analysis of categorical data - is Chi-Square my only
option?
You are absolutely correct - my vocabulary may be my undoing here. I
guess the ultimately question is, when looking at my population, I
want to see what variables have a significant relationship (or effect)
with the "quit" variable. That is why at first I was using
Chi-Squares, but since I am looking at 6 or so influencing the "quit"
variable, it gets cumbersome.
I was wondering (hoping) if there was a "kitchen sink" approach that
throws all of them in the pot and comes back with what matters and
what does not.
Thanks for the help.
On 11/17/05, Aric Zion <Aric.Zion@asu.edu> wrote:
> You said that your IV is the quit variable. Do you mean instead that
your DV (i.e. the thing you want to predict) is the quit variable?
>
> You could also use logistic regression to look at your problem. That
would also give you significance tests for each category of each of
your IVs (i.e. predictors). Of course, if use logistic regression, you
will have to decide which category is your reference category for each
of your IVs.
>
> Aric
>
> -----Original Message-----
> From: SPSSX(r) Discussion on behalf of E. Griffin
> Sent: Thu 11/17/2005 12:18 PM
> To: SPSSX-L@LISTSERV.UGA.EDU
> Cc:
> Subject: Analysis of categorical data - is Chi-Square my only option?
>
>
>
> I have tried to search for an answer, but need to appeal to all
for
> help. I have a 10,000 case data set consisting of demographic
data in
> categorical format. The data set is for a group of people, and
the IV
> I ultimately am looking at is whether they stay in the group or
quit.
> I have this coded as a variable - a '1' if they quit and a '0'
if they
> did not.
>
> I have run crosstabs to look at relationships between
demographic data
> (gender, ethnicity, etc.) and this "Quit" variable and produce
a
> Chi-Square. Is this the best way to look at this, or is there
a
> better way?
>
> Since I am looking at perhaps 5 or 6 possible categorical DV I
feel
> like maybe I am going about this the wrong way. Thanks for any
help.
>
>
>
>
| 
