ICPR Abstracts: Session 29

---

Session 29: Symposium

Analyzing Data From Couples:
Techniques for Preserving Interdependence


The Dangers of Difference Scores in Dyadic Data

Dale Griffin
University of Sussex

There are many situations in which it is tempting to use 
difference scores to index "discrepancy" or "agreement" 
between members of couples.  In most situations, this 
temptation should be resisted or the researcher may be led 
into seriously mistaken conclusions.  The classic criticisms 
of difference scores (e.g., unreliability and bias) will be 
reviewed and then will be translated into relationship 
terms to illustrate the general nature of the problem. Three 
common research paradigms will be discussed (agreement 
between partners, comparison of partners with ideal 
standards, and the accuracy of individuals' self-ratings) 
and the (mistaken) conclusions generated by analysis of 
difference scores will be contrasted with the conclusions 
generated by the (appropriate) analysis using regression 
methods. 


Correlational Methods for Distinguishable Dyads

Richard Gonzalez
University of Washington

This talk focuses on the appropriate analysis of 
correlational designs when each dyad is made up of two 
different types of individuals (e.g., a heterosexual couple 
or a parent and a child). Too often, researchers separate 
their interdependent samples into two different samples of 
independent observations (e.g., computing correlations for 
women and for men separately).  This talk describes the 
three stages of analysis appropriate for this design: first, 
the interdependent nature of the sample is modeled and 
assumptions of equal correlations for the two types of 
individuals are tested; second, the overall correlation 
across all members is assessed with a test that takes the 
degree of interdependence into account; finally, 
correlations at two different levels of analysis (dyadic 
level and individual level) can be computed and 
interpreted.  Each stage of analysis is illustrated with data 
from studies of heterosexual couples.  Finally, the 
conceptual meaning of individual versus dyad-level 
analysis is discussed along with an explanation of why 
dyad mean scores do not represent "dyad-level" processes. 


The Challenges of Dyadic Data Analysis in Practice:
An Editor's Viewpoint

John Holmes
University of Waterloo

In this talk, I review some of the most perplexing data 
analysis issues that relationship researchers typically face.  
In addition to reflecting upon the perspectives introduced 
by the methodological talks, I intend to point out those 
particular challenges in dyadic analysis that have not been 
solved clearly enough for the average researcher. 


Using Structural Equation Modeling to Examine Sex 
Differences in Relationship Processes

Sandra Murray
University of Michigan

Many research designs use regression or path models to 
explore the relations among a set of variables.  In studying 
heterosexual relationships, a researcher faces the problem 
of modeling the relations between the two types of 
individuals within each couple as well as modeling the 
relations between variables within each individual.  In this 
talk, I explain the use of structural equation modeling in 
such situations and give several examples from a study of 
"positive illusions" in romantic relationships.  In 
particular, I illustrate the value of using SEM for 
identifying and testing sex differences in path models and 
contrast it with the limits of standard regression methods.  
The critical role of a priori theory in SEM will be stressed 
and differences between this path model approach and the 
use of SEM for testing "latent variable" models will also 
be discussed.