SPSS-Need help creating a variable for dyadic data

[JarvisCain]

I have data for 200+ couples, one row per individual. Each couple shares an identifier (their CoupleID) which occupies 1 column. Each couple answered a set of questions about themselves, and the same questions about their partner. Individuals without partners have a blank CoupleID.

I need to create a new variable that correlates each respondent's answers about their partner with that partner's answers about themselves. So if you look at my example, Couple 1 (cplID=1) would correlate perfectly with each other, as each one accurately predicted the other's answer on all four items. For Couple 2 though, only one partner (ID#=41353) predicted their partner's responses perfectly, while that partner's predictions were a little off on items 2 and 3, and way off on Item 4.

So, I need to tell SPSS to create a variable that looks at each pair by their CoupleID and compares the Partner response for each one to the Self response of the other.

Thank you!

 

[Karabiner]

Is it really correlation you are looking for? If partner A answers 1 2 3 4, and partner B answers 3 4 5 6, then this would
be a perfect correlation (r=1), but obviously the partners disagree a lot.

With kind regards

Karabiner

 

[JarvisCain]

Yes, what I'm interested in is the strength of correlation between Partner A's prediction for Partner B's responses (and vice versa) for each item, if that helps clarify it.

I have a lot of other demographic variables, like relationship duration, gender, etc., so that I can describe patterns like "For Item 2, couples who have been together longer have much higher accuracy for predicting their partners' responses than new couples" or perhaps "For Item 3, women were significantly better at predicting their partner's responses than men were." I don't expect any perfect correlations, of course, I'm interested in what predicts levels of accuracy.

 

[Karabiner]

So you want to analyse each item separately? And for each item, you want to predict the difference
between self-assessement and assessment-by-partner?

With kind regards

Karabiner

 

[JarvisCain]

Yes, I want to be able to analyse each one seperately, but to also be able group items into composite scales.

The actual measure is a 60-item five-factor model personality instrument, with 2 questions for each of the 30 facets, and with 6 facets per factor being measured for each partner. So, I'll need to analyse at the factor level and the facet level. The idea is that some aspects of personality become evident to partners very early, while others may take years to either develop or be expressed.

Thanks

 

[Karabiner]

So if item1 and item2 are one facet, you calculate

disagreement_facet1=ABS( (item1_self+item2_self)-(item1_partner+item2_partner) ).

You can use this absolute difference score as dependent variable in e.g. a linear regression analysis.

With kind regards

Karabiner

 

[JarvisCain]

Well, the items have already been combined from their composite questions, but is this the script to use for creating an absolute value variable in SPSS? What I've done is create a disagreement variable that subtracts the partner actual from the respondent's prediction, which sometimes results in a negative integer, so I thought I may have to square the differences to calculate the variance.