Purpose

This workflow estimates an actor-partner interdependence model (APIM) for distinguishable dyads. It evaluates whether each person’s stress predicts their own satisfaction (actor effect) and the other member’s satisfaction (partner effect) while modeling residual dependence within dyads.

The included dyadic data are entirely synthetic. Two observations from a dyad are not independent. Define who belongs to each dyad, whether members are meaningfully distinguishable, and how actor and partner variables are aligned before fitting the model.

APIM concepts

  • Actor effect: association between a person’s predictor and their own outcome.
  • Partner effect: association between a person’s predictor and the other member’s outcome.
  • Dyadic nonindependence: outcomes or errors are correlated within a pair.
  • Distinguishable dyads: members have a meaningful role that can be assigned consistently, such as doctoral student and support partner.
  • Indistinguishable dyads: roles cannot be assigned meaningfully; these require equality constraints and different inferential handling.

An APIM does not establish interpersonal causation without a design that supports causal inference.

User settings

Simulate and import the example

Rows: 300
Columns: 6
$ dyad_id              <chr> "D001", "D002", "D003", "D004", "D005", "D006", "~
$ relationship_years   <dbl> 1.5, 5.2, 5.7, 5.8, 7.8, 3.9, 6.4, 9.8, 18.4, 17.~
$ student_stress       <dbl> 36.7, 49.8, 54.4, 63.6, 27.7, 47.2, 38.7, 40.5, 6~
$ partner_stress       <dbl> 40.5, 35.7, 36.0, 37.4, 36.2, 56.4, 28.3, 57.7, 4~
$ student_satisfaction <dbl> 86.6, 91.3, 73.9, 65.9, 100.0, 81.2, 85.9, 80.9, ~
$ partner_satisfaction <dbl> 82.2, 86.7, 74.2, 57.9, 96.5, 77.3, 74.2, 73.2, 6~

Audit the dyadic structure

dyads unique_dyad_ids complete_dyads student_outcomes_observed partner_outcomes_observed
300 300 276 288 288

Every row in the wide file represents one dyad. Member-specific variables must be aligned consistently; a swapped actor-partner column changes the scientific meaning of the model.

Describe dyadic association

student_stress partner_stress student_satisfaction partner_satisfaction
student_stress 1.00 0.36 -0.64 -0.53
partner_stress 0.36 1.00 -0.46 -0.58
student_satisfaction -0.64 -0.46 1.00 0.67
partner_satisfaction -0.53 -0.58 0.67 1.00

These correlations describe nonindependence but do not separate actor and partner associations.

Prepare the two-row-per-dyad data

n dyads
2 300

The actor and partner columns are person-centered by role only in their meaning, not statistically centered within dyad. They are grand-mean centered across the two-row dataset so the role-specific intercepts describe an average-stress dyad.

Fit the distinguishable-dyad APIM

estimation_rows estimation_dyads complete_estimation_dyads partial_estimation_dyads
576 300 276 24
term estimate std_error t_value p_value conf_low conf_high
roleStudent 73.108 0.593 123.354 0.000 71.944 74.273
rolePartner 69.594 0.615 113.246 0.000 68.387 70.802
relationship_years 0.135 0.048 2.799 0.005 0.040 0.229
roleStudent:actor_stress_c -0.519 0.044 -11.922 0.000 -0.605 -0.434
rolePartner:actor_stress_c -0.486 0.051 -9.575 0.000 -0.586 -0.387
roleStudent:partner_stress_c -0.279 0.048 -5.868 0.000 -0.372 -0.186
rolePartner:partner_stress_c -0.373 0.046 -8.099 0.000 -0.464 -0.283

The no-intercept parameterization estimates a separate intercept, actor effect, and partner effect for each role. The compound-symmetric residual structure models the remaining correlation between the two outcomes in a dyad, and varIdent allows role-specific residual variances.

Summarize actor and partner effects

effect estimate std_error t_value p_value conf_low conf_high
Student actor effect -0.519 0.044 -11.922 0.000 -0.605 -0.434
Student partner effect -0.279 0.048 -5.868 0.000 -0.372 -0.186
Partner actor effect -0.486 0.051 -9.575 0.000 -0.586 -0.387
Partner partner effect -0.373 0.046 -8.099 0.000 -0.464 -0.283
Difference in actor effects -0.033 0.071 -0.462 0.644 -0.173 0.107
Difference in partner effects 0.094 0.071 1.334 0.183 -0.045 0.233

The role-difference contrasts test whether actor or partner associations differ between the doctoral student and support partner. A significant effect for one role and nonsignificant effect for the other does not itself establish a significant role difference.

Predicted actor associations

Model diagnostics

Review linearity, role-specific residual distributions, influential dyads, missingness, actor-partner alignment, and whether the assumed within-dyad correlation is adequate. Longitudinal dyadic data need an additional time level and a model that separates within-person, between-person, and dyadic dependence.

Dynamic results template

A distinguishable-dyad actor-partner interdependence model was fitted to 576 observed outcomes from 300 synthetic dyads (276 complete and 24 partially observed in the estimation sample). The residual correlation between paired outcomes was 0.4 after accounting for the predictors.

For doctoral students, a one-point increase in their own stress was associated with a -0.52-point difference in their satisfaction, 95% CI [-0.6, -0.43], p < .001. A one-point increase in the support partner’s stress was associated with a -0.28-point difference in student satisfaction, 95% CI [-0.37, -0.19], p < .001.

For support partners, the actor effect was -0.49, 95% CI [-0.59, -0.39], p < .001, and the partner effect of student stress was -0.37, 95% CI [-0.46, -0.28], p < .001. Interpret actor and partner associations as conditional relationships rather than causal effects.

Reporting checklist

  • Define the dyad, roles, distinguishability decision, and alignment procedure.
  • State the number of complete and partially observed dyads.
  • Define actor and partner predictors and their centering.
  • Report role-specific actor and partner effects with confidence intervals.
  • Test role differences directly rather than comparing significance labels.
  • Describe the within-dyad correlation and role-specific residual variance structure.
  • Document diagnostics, missing-data assumptions, and limits on causal interpretation.

Exports

file
synthetic_apim_dyads_wide.csv
synthetic_apim_dyads_long.csv
apim_coefficients.csv
actor_partner_effects.csv
dyadic_correlations.csv
model_sample.csv
predicted_actor_associations.csv
run_settings.csv
package_versions.csv

References

  • Kenny, D. A., Kashy, D. A., & Cook, W. L. (2006). Dyadic Data Analysis. Guilford Press.
  • Ledermann, T., Macho, S., & Kenny, D. A. (2011). Assessing mediation in dyadic data using the actor-partner interdependence model. Structural Equation Modeling, 18(4), 595–612.
Dissertation Stats Helper · Reproducible statistical guidance for doctoral research · Preserve dyad membership, role alignment, and within-pair dependence.