15.2 Interdependence
One of the core concepts in dyadic data analysis is interdependence.
15.2.1 Definition
In the context of dyads, we consider two members of a dyad to be interdependent (or linked) if the data points collected from the dyad are
- more similar than one might expect from two individuals not in that dyad, or
- more dissimilar than one might expect from two individuals not in that dyad.
15.2.2 Ignoring Interdependence
Ignoring the interdependence in dyad members’ scores distorts standard significance tests.
This distortion arises primarily through biases in variances. Their is also a loss in the degrees of freedom due to the potential redundancy of information among dependent pairs. Consider the extreme case of two identical dyad members.
For example, if the correlation between dyad members’ scores is positive, the estimated variance of the observations would be smaller than it should be if all individuals were independent.
If the correlation were negative, the variances would be larger than they should be.
15.2.3 Linkage Types
Interdependence, or linked scores, can occur in several ways, including the a few common, potentially overlapping, scenarios:
- Voluntary linkage: the dependence that arises through a relationship over time. Typical examples are linkages between friends, between members of dating couples.
- Kinship linkage: the dependence that arises through family relations. Typical examples include family members, siblings, cousins, parents and children.
- Experimental linkage: the dependence that arises from an artificially contrived relationship for the purpose of study. Typically examples include the prompt for two subjects to get to know eachother.
- Yoked linkgage: the dependence that arises not from interpersonal interactions, but from two subjects being exposed to the same environmental stimuli.
15.2.4 Sources of Interdependence
Furthermore, Kenny (1996) and Kenny, Kashy and Cook (2006) elaborate on four sources that may generate interdependence in dyads.
1. Compositional Effects
Compositional effects will often occur in naturally occurring dyads, or anytime dyad members are paired together in a nonrandom way. These are typically factors that induce similarities among dyad members prior to the forming of the dyad.
For example, compositional effects are to be expected with friends, even before they meet, as friends are typically similar on a wide range of variables, including education, age, socioeconomic status, religion, and many other characteristics. Random pairing of a dyad may alleviate certain types of compositional effects.
2. A Partner Effect
A partner effect occurs when the behavior (or certain characteristics) of one dyad member affects the behavior (or outcomes) of the partner dyad member. Typically partner effects occur after the instantiation of the dyad.
For example, the amount of housework that one roommate does may affect the other roommate’s satisfaction level in regard to their living arrangements. This could be considered a partner effect.
3. Mutual Influece
Mutual influence occurs when both persons’ outcomes directly affect one another. Typically, mutual influence stipulates a feedback process of some kind.
For example, suppose a research is interested in measuring how much two experimentally linked dyad members like eachoter. How much one person likes their interaction partner can directly impact how much the other partner likes them in return.
4. Common Fate
Common fate effects occur when both dyad members are exposed to the same causal factors.
For example, if two children are exposed to the same neglect growing up, their parental relationships may be similar due to common experiences during childhood.