New variable as a combination of responses

[Koen Bruelemans]

I'm pretty new at SPSS and I've encountered the following problem.

I have a dataset where the behaviour of every time two participants to an
interaction was coded. In the dataset, a lot of variables are mentioned
twice (for every participant). For example: SexPart1 for the sex of
participant 1, SexPart2 for the sex of participant 1, EtnPart1  for the
etnicity of participant 1, EntPart2 for the etnicity of participant 2, and
so on.

I have already made multiple response sets for calculating frequencies and
percentages for all participants, but that doesn't allow me to calculate
chi-square and correlation. I already found out that I have to create new
variables where every value is a combination of the responses of the old
variables (for example '1, 2' if participant 1 is male and participant 2 is
female). I have to do this with the compute command, but can someone explain
me with an example in detail how exactly I have to go about this?

Thank you.

[Richard Ristow]

At 01:51 PM 4/29/2007, Koen Bruelemans wrote:

>I have to create new variables where every value is a combination of
>the responses of the old variables (for example '1, 2' if participant
>1 is male [SexPart1] and participant 2 is female [SexPart2]).

You'll have to create a string variable. Assuming that SexPart1 and
SexPart2 are numeric, and have one relevant digit each, this (untested)
should do what you want:

STRING  SexBoth (A3).
COMPUTE SexBoth = CONCAT(String(SexPart1,F1),
                          ',',
                          String(SexPart2,F1)).

Now, what you're doing raises some questions for me, and I'll leave you
with those. They may or may not be of use, but I'm not comfortable
leaving that solution as final.

First, is there a meaningful distinction between which participant is
'participant 1' and which is 'participant 2'? So, that is, is there a
meaningful difference between '1,2' and '2,1' - participant 1 male and
2 female, vs. the reverse? Or should the classification be, simply,
"both male", "both female", and "different"?

>I have already made multiple response sets for calculating frequencies
>and percentages for all participants, but that doesn't allow me to
>calculate chi-square and correlation.

Would you say what you want to test by cross-tabulation (I assume
that's where your chi-square will come from) and correlation? What,
indeed, are you studying altogether? Comparing members of each pair,
comparing pairs with other pairs? There may be alternative ways of
looking at your data, and at your problem, and some on this list may
have suggestions.

I hope what I first suggested solves your problem as you gave it; and I
hope your study goes well, overall.

Good luck to you,
Richard Ristow