# How to do 2x2, between and within, analysis?

Discussion in 'SPSS' started by Jim Kroger, Nov 18, 2004.

1. ### Jim KrogerGuest

Hi. I've always done within subjects designs, and now need to analyze
a design with one dependent variable measured across two groups.
Musicians and non-musicians were each given both a spatil reasoning
test and a verbal reasoning test. The design sort of looks like the
following, in which musical ability is a between subjects factor,
and reasoning is a within subjects factor.

musical abillity
musician nonmusician

spatial
reasoning
verbal

I'm having trouble figuring out how arrange the data, and how to
do the test. Is it the univariate GLM? I presume reasoning is the
random factor and musical ability if the fixed factor. But I'm unsure
how to handle the dependent variable? Does it appear as a separate
column in the data sheet?

Thanks in advance for any help or pointers to info.

Jim

Jim Kroger, Nov 18, 2004

2. ### Bruce WeaverGuest

Your file should have one row per person with (at least) these 3
columns: One to code musician v non-musician; and one for each
reasoning score.
See above.
Repeated measures ANOVA can used when you measure the *same* variable
two or more times for each person. It is not meant for the case where
measure two or more *different* variables for each person. In that
case, you might use MANOVA.

So, are spatial and verbal reasoning two measures of the same thing? Or
are they two different things? (FWIW, I think they are two different
things.)

To get SPSS to do the work...

For repeated measures: Analyze/GLM/Repeated measures
For MANOVA: Analyze/GLM/Multivariate

To save the syntax, exit the dialogs via <Paste> rather than <Okay>.

The MANOVA will give you so-called univariate tests, which for your data
would be equivalent to two unpaired t-tests, one for each type of
reasoning. Given that you have only two measures, it might be simpler
to just go with these two unpaired t-tests.

Bruce Weaver, Nov 18, 2004

3. ### Jim KrogerGuest

Bruce, thank you so much for your help. I tried it, and it all works
great with what the command interface wanted, and the analysis looks
good.

Jim

Jim Kroger, Nov 19, 2004
4. ### Bruce WeaverGuest

Grand. Just out of curiousity, did you decide to treat the two
reasoning scores as repeated measures of the same variable, or as two
different variables?

Bruce Weaver, Nov 19, 2004
5. ### Jim KrogerGuest

Hi Bruce. Well, we decided to treat it as two levels of the variable
reasoning. You've probably heard of the Mozart effect, which improves
spatial reasoning. We're comparing its effect on verbal reasoning. So
the critical issue is comparison of spatial to verbal reasoning, so it
seemed intuitive to treat them as levels of a variable.