Re: CATREG for ordinal x continuous

[Art Kendall-2]

I have only read some of the CATREG documentation, so do not feel
qualified to answer.

I had the impression that there was "automatic" testing of levels of
measurement assumptions.

I suggest going through the archives of this list to find the names of
the people from Leiden who created the CATEGORIES module and addressing
the question to them.

Art Kendall
Social Research Consultants


Dale Glaser wrote:

> Art, you were kind enough to respond to my ordinal x continuous level
> variable inquiry for a regression model...so I did update my
> CATEGORIES module today (when your self-employed these modules don't
> come cheap!!) and I noticed that it does not have an interface similar
> to logistic regression where you can create the multiplicative terms.
> Thus, it doesn't make sense to identify one continuous
> predictor variable in CATREG as numeric, the ordinal predictor as
> ordinal and create the multplicative term as I would in OLS.  However
> Art, I was wondering how kosher do you think it would be to use the
> discertize function in CATREG after I assign the ordinal spline (2 2)
> transformation to the ordinal IV, and then since I notice SPSS saves
> this transformed variable, create the multplicative interaction term
> and run this in CATREG?.........................or am I way off
> base?!!.....dale
>
> */Dale Glaser <[hidden email]>/* wrote:
>
>     Hi Art..I just re-examined the data and it is # of hours
>     telecommuting per week (the focus of the study)......and I haven't
>     looked at CATREG, though that may be an option....thank you....dale
>
>     */Art Kendall <[hidden email]>/* wrote:
>
>         Did you look at CATREG -- categorical regression? I have the
>         impression
>         that it would test the fit of different levels of measurement.
>
>         Just curious. Is your commuting time hours per month? Or is
>         this some
>         very unusual pop?
>
>         Art Kendall
>         Social Research Consultants
>
>         Dale Glaser wrote:
>         > I am on a project where I will be testing an interaction
>         between a continous level variable and what is in practice a
>         continuous level moderator (hours commute); however, 'hours
>         commute' was captured as an ordered categorical variable (1 =
>         < 5 hrs; 2 = 5-10; 3 = 11-15...........to 8 = > 40 hours);
>         thus, to capture the interaction in a moderated multiple
>         regression (MMR) context, I usually would heed the advice of
>         Cohen et al (2003) and center the predictors so as to decrease
>         collinearity of the lower order terms. However, with an
>         ordinal variable, even if the scaling is deemed to be
>         arbitrary, it seems problematic to mean center such a
>         variables as well as create the multiplicative term. So I was
>         curious what strategy any of you employ when creating an
>         ordinal x continuous level interaction term in a MMR context.
>         I know that amongst the strategies used in structural equation
>         modeling (SEM) for models with moderators, one does
>         incorporate creating multiplicative terms
>         > between the manifest indicators for the latent constructs
>         (e.g, using LISREL notation, LX21 x LX22 for the 2nd item
>         loading on the first two latent constructs) and often those
>         are ordinal, self-report items, and centering may or may not
>         be executed. So, off the top of my head here is my proposed
>         options, and I would be most appreciative to solicit any of
>         your opinions:
>         >
>         > (1) assume that this is much ballyhoo about nothing and
>         create the continuous x ordinal multiplicative term with
>         impunity (and centering is fine under the assumption of
>         arbitrariness of scaling for the ordinal variable), though it
>         would seem caution is in order for interpreting the
>         unstandardized partial regression coefficient
>         >
>         > (2) since 64% of the sample from this project commute > 40
>         hours, create a dummy coded binary variable coding for '< 40
>         hours' and '> 40 hours', but lose the rank-ordering nature of
>         the variable and attendant information (leading to truncation
>         of variation).
>         >
>         > (3) and less appealing,create 7 dummy coded vectors (to
>         capture the 8 levels of 'hours') and create a potentially
>         over-parameterized model with the continous x 7 dummy coded
>         vectors, and as with option #2 lose the theoretical continuity
>         of the moderator variable
>         >
>         > (4) I was trying to think if there was strategy akin to
>         polyserial/polychoric correlation where I could create some
>         type of thresold parameter for the ordinal variable, but I'm
>         not sure of the advisability of such an approach.
>         >
>         > Any feedback would be most appreciate...thank you....
>         >
>         > Dale Glaser
>         >
>         >
>         > Dale Glaser, Ph.D.
>         > Principal--Glaser Consulting
>         > Lecturer/Adjunct Faculty--SDSU/USD/AIU
>         > President-Elect, San Diego Chapter of
>         > American Statistical Association
>         > 3115 4th Avenue
>         > San Diego, CA 92103
>         > phone: [phone]
>         > fax: [phone]
>         > email: [hidden email]
>         > website: www.glaserconsult.com
>         >
>         >
>         >
>
>
>
>
>
>     Dale Glaser, Ph.D.
>     Principal--Glaser Consulting
>     Lecturer/Adjunct Faculty--SDSU/USD/AIU
>     President-Elect, San Diego Chapter of
>     American Statistical Association
>     3115 4th Avenue
>     San Diego, CA 92103
>     phone: [phone]
>     fax: [phone]
>     email: [hidden email]
>     website: www.glaserconsult.com
>
>
>
>
> Dale Glaser, Ph.D.
> Principal--Glaser Consulting
> Lecturer/Adjunct Faculty--SDSU/USD/AIU
> President-Elect, San Diego Chapter of
> American Statistical Association
> 3115 4th Avenue
> San Diego, CA 92103
> phone: [phone]
> fax: [phone]
> email: [hidden email]
> website: www.glaserconsult.com

[aysemel]

Hi everyone,

i have a very very urgent situation and i would be very glad if someone can help me.  I should do a ridge regression analyis for continuous variables on SPSS 20, how can i do that? by using CATREG menu? how can i interpret this? Could you please explain me the method and the steps that i should do. How can i compare the results that i have from the linear regression and ridge regression? Can i get unstandardized coefficients for ridge regression? Is there a way for these issues other than writing a syntax in SPSS?  

Yes, the RR model is under Analyze>Regression>Optimal Scaling



On 6/6/2012 1:47:47 PM, aysemel ([hidden email]) wrote:

> Hi everyone,
>
> i have a very very urgent situation and i would be very glad if someone
> can
> help me.  I should do a ridge regression analyis for continuous
> variables on
> SPSS 20, how can i do that? by using CATREG menu? how can i interpret
> this?
> Could you please explain me the method and the steps that i should do.
> How
> can i compare the results that i have from the linear regression and
> ridge
> regression? Can i get unstandardized coefficients for ridge regression?
> Is
> there a way for these issues other than writing a syntax in SPSS?
>
> --
> View this message in context: http://spssx-discussion.1045642.n5.nabble.
> com/Re-CATREG-for-ordinal-x-continuous-tp1074799p5713543.html
> Sent from the SPSSX Discussion mailing list archive at Nabble.com.
>
> =====================
> 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

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

[aysemel]

Thanks WillBaileyz.

Do you have an idea abou thow can i interpret the results? how can i compare the results that i have from the linear regression and  ridge regression? Can i get unstandardized coefficients for ridge regression on SPSS?