Google Groups no longer supports new Usenet posts or subscriptions. Historical content remains viewable.
Dismiss

How to explore relations between categorical variables?

3 views
Skip to first unread message

hamha

unread,
Dec 11, 2009, 5:45:33 AM12/11/09
to
Hi,
As I am fairly unexperienced, I need some advice. To my great
frustration, I have spent to days approaching an idea, without getting
anywhere ;-)
In short, this is what I hope to achieve:
1) I have 5 categorical variables, with 4 categories (values) on each
variable.
2) Accordingly, it is a total of 20 categories.
3) I want to see to what extent each individual (n= 10546) have
corresponding values on these different variables. For instance, I
want to check if those who agree with claim nr 1 do also agree with
claim nr 2.
4) Preferably, I was hoping to display this correspondence
graphically.
(I should perhaps mention that I am used to exploring this kind of
structures by way of multiple correspondence analysis – in the program
SPAD – but I seem unable to do similar operations in SPSS.

All suggestions would be greatly appreciated!


Bruce Weaver

unread,
Dec 11, 2009, 8:55:05 AM12/11/09
to

How about:

Model Selection Loglinear Analysis
The Model Selection Loglinear Analysis procedure analyzes multiway
crosstabulations (contingency tables). It fits hierarchical loglinear
models to multidimensional crosstabulations using an iterative
proportional-fitting algorithm. This procedure helps you find out
which categorical variables are associated. To build models, forced
entry and backward elimination methods are available. For saturated
models, you can request parameter estimates and tests of partial
association. A saturated model adds 0.5 to all cells.

Example. In a study of user preference for one of two laundry
detergents, researchers counted people in each group, combining
various categories of water softness (soft, medium, or hard), previous
use of one of the brands, and washing temperature (cold or hot). They
found how temperature is related to water softness and also to brand
preference.

Statistics. Frequencies, residuals, parameter estimates, standard
errors, confidence intervals, and tests of partial association. For
custom models, plots of residuals and normal probability plots.

In the GUI: Analyze - Loglinear - Model Selection

HTH.

--
Bruce Weaver
bwe...@lakeheadu.ca
http://sites.google.com/a/lakeheadu.ca/bweaver/Home
"When all else fails, RTFM."

Ray Koopman

unread,
Dec 11, 2009, 1:53:27 PM12/11/09
to

Are the categories ordered? Are they the same for all 5 variables?

Rich Ulrich

unread,
Dec 11, 2009, 4:55:27 PM12/11/09
to

The results? If the variables were independent, the Model
indicated would be the frequencies on 5 separate variables.
That seldom happens.

When the whole set of dependencies can be explained by
the 2-way tables, then the results show you that 3-way
and higher are not needed. That is not uncommon -- but
with N of 10 000, there may be so much "statistical power"
available, that you will get tests that suggest that 3-way
tables, etc., are needed to explain all the results.

- Ray raises a point about "ordering": if the categories
are ordered (bad to good; least to most), the tests that
assume unordered categories are not the most sensitive.

- SPSS does have something called correspondence
analysis, which I've used. The description above seems
like a simpler design than what I remember using it for.
But you might look for it and check.

--
Rich Ulrich

0 new messages