Report

Patterns of Actor and Partner Effects David A. Kenny February 17, 2013 You need to know the Actor Partner Interdependence Model! APIM 2 APIM Patterns: Couple Model • Model – Equal actor and partner effects: a = p – e.g., my depressive symptoms has the same effect on my quality of life as does my partner’s depressive symptoms on my quality of life • Average or sum as the predictor – Although measured individually, the predictor variable is a “dyadic” variable, not an individual one 3 APIM Patterns: Contrast • Model – Actor plus partner effects equals zero: a – p = 0 – Klumb et al. (2006): time spent doing household labor on stress levels • The more household labor I do, the more stressed I feel. • The more household labor my partner does, the less stress I feel. • Difference score (actor X minus partner X) as the predictor 4 APIM Patterns: Actor or Partner Only • Actor Only – Actor present but no partner effect – Fix the partner effect to zero. • Partner Only – Partner present but no partner effect – Fix the actor effect to zero. – Relatively rare. 5 Testing Patterns • Multilevel Modeling – Sum and difference approach • Structural Equation Modeling – Setting coefficients equal – Use of phantom variables • General approach to patterns: k 6 Sum and Difference Approach • Remove the actor and partner variables from the model. • Add to the model the Sum and the Difference score as predictors. • If Sum is present, but not the Difference, you have a couple model. • If Sum is not present, but the Difference is, you have a contrast model. 7 Acitelli Example • Distinguishable – Husbands • Sum: 0.392, p < .001 • Difference: 0.131, p = .088 – Wives • Sum: 0.373, p < .001 • Difference: 0.001, p = .986 • Indistinguishable – Sum: 0.344, p < .001 – Difference: 0.056, p = .052 8 Testing the Couple Model Using SEM • Actor effect equal to the partner effect. • Can be done by setting paths equal. • Distinguishable dyads a1 = p12 and a2 = p21 • Indistinguishable dyads a=p 9 Acitelli Example • Distinguishable –Husbands: 0.346 –Wives: 0.347 –Test: c2(2) = 4.491, p = .106 • Indistinguishable –Effect: 0.344 –Test: c2(1) = 3.803, p = .051 10 Testing the Contrast Model Using SEM • Actor effect equal to the partner effect times minus 1. • Can be done by using a phantom variable. • Phantom variable – No conceptual meaning – Forces a constraint – Latent variable – No disturbance 11 Contrast Constraint Forced by Phantom Variables (P1 and P2) a1 X1 Y1 1 E 1 -1 a2 P1 a1 P2 -1 X2 a2 Y2 1 E2 • Now the indirect effect from X2 to Y1, p12 equals (-1)a1 12 Acitelli Example c2(2) = 69.791, p < .001 13 Conclusion Using patterns can link the APIM to theory and simplify the model. The k parameter is a general way to measure and test patterns Readings pp. 147-149, in Dyadic Data Analysis by Kenny, Kashy, and Cook Kenny & Cook, (1999), Personal Relationships, 6, pp. 433-448. 14