Moderated Actor-Partner Interdependence Model

The Moderated Actor-Partner Interdependence Model (APIM-Mod)

While the basic APIM allows us to estimate the direct influence of one person's variables on another's (partner effects), it assumes these effects are constant for everyone. Moderation analysis relaxes this assumption. It formally tests whether the strength or direction of an actor or partner effect is conditional on the value of a third variable, the moderator. This is statistically represented by an interaction term.

A foundational paper by Chow et al. provides an excellent primer on integrating moderation into the APIM framework. It organizes moderators into three distinct types, which is a useful way to conceptualize their role in a dyadic or triadic system.

The "mixed moderator" concept is precisely what we need to address today's learning outcome. To test if the bull's sensitivity moderates the effect of a specific activity on the wife's guilt, we would model an interaction. The bull's sensitivity is the moderator, and it's a "mixed" variable because it varies from bull to bull (between dyads) and is distinct from the wife's characteristics (within the dyad). The model would test whether the path from Activity to Wife's Guilt changes depending on the level of Bull's Sensitivity.

Specifying and Interpreting a Moderated Triadic Model: A Case Study

Having established the concept, let's examine a sophisticated application. The following study uses a triadic APIM and tests for moderation to understand complex family dynamics. While the context is dual-earner parents and their adolescent children, the statistical methodology is directly transferable to your husband-wife-bull triad.

The authors use a technique called multi-group analysis, which is a powerful way to test for moderation when the moderator is a categorical variable (e.g., employed vs. self-employed). The logic involves splitting the sample based on the moderator and comparing the strength of the causal paths between the groups. A significant difference indicates moderation.

Applying Moderation to Your Research Scenario

Let's translate this methodology to your work. Suppose you want to test the hypothesis: "The bull's attachment style (secure vs. insecure) moderates the effect of the wife participating in a high-intimacy activity on the husband's jealousy."

Here, the causal path of interest is Activity → Husband's Jealousy (a partner effect), and the moderator is the Bull's Attachment Style (a categorical variable).

Following the logic from the Schnettler et al. paper, your analysis plan would be:

1. Group Your Data: Divide your sample of triads into two groups: those with a "secure" bull and those with an "insecure" bull.
2. Specify the APIM: Define an APIM for the husband-wife dyad that includes the Activity variable as a predictor of both actor (wife's) and partner (husband's) outcomes.
3. Run a Multi-Group SEM:
   • Fit the APIM simultaneously for both groups.
   • Initially, constrain the partner path from Activity → Husband's Jealousy to be equal across the two groups.
   • Then, fit a second model where you allow this path coefficient to be different for the "secure bull" group and the "insecure bull" group.
4. Test for Moderation: Compare the fit of the constrained and unconstrained models using a chi-square difference test ($\Delta {\chi }^2$). A statistically significant result indicates that the path is indeed different between the two groups—in other words, the bull's attachment style moderates the activity's effect on the husband's jealousy.

What if your moderator is continuous, like the bull's score on a sensitivity scale? In that case, you would revert to the interaction term approach described in the Chow et al. paper. Within your SEM, you would specify the husband's jealousy as a function of:

• The Activity
• The Bull's Sensitivity Score
• The interaction term: Activity × Bull's Sensitivity Score

A significant path coefficient for the interaction term would be your evidence for moderation. You can then probe this interaction to understand its nature—for example, does the activity only increase jealousy when the bull has low sensitivity?

Conclusion

In this lesson, we have moved beyond simple direct effects to explore the conditional nature of causal relationships in triads. By incorporating moderation into the Actor-Partner Interdependence Model, you can now specify models that test when and for whom certain activities or characteristics have their strongest impact.

Key Takeaways:

• Moderated APIMs test whether actor or partner effects are conditional on a third variable (the moderator). This is statistically evaluated using an interaction effect.
• Moderators can be characteristics of individuals (e.g., personality), dyads (e.g., roles), or entire triads (e.g., relationship rules).
• For categorical moderators, a powerful specification method is multi-group SEM, where you compare path strengths across groups.
• For continuous moderators, you specify the model by creating a product term (interaction term) between the predictor and the moderator.
• The ultimate goal is to move from a significant interaction term to a rich causal story explaining the conditions under which an effect occurs.

In our next lesson, we will focus on the practical details of this final step. We will cover how to estimate and interpret the interaction term representing cross-person moderation, including the essential techniques of probing and plotting simple slopes to visualize and explain precisely what the moderation means.