Path analysis was implemented to examine pathways from couple relationships to adolescent internalizing problems. Although different analytic approaches could be used to evaluate components of the proposed model, by simultaneously estimating all possible associations path modeling conservatively estimates model parameters. Critical to our different pathways hypotheses is that indirect effects can also be estimated. Models were estimated using Mplus Version 6.1 (Muthén & Muthén, 1998–2012). Full-information maximum-likelihood estimation was used to reduce potential bias incurred due to missing data at later waves (see Schlomer, Bauman, & Card, 2010). Analyses were conducted by first examining overall model fit indices, chi-square, the comparative fit index (CFI), the nonnormed or Tucker–Lewis index (TLI), and the root-mean-square error of approximation (RMSEA) and its standardized root-mean-square residual (SRMR). Models met criteria for adequate overall fit when CFI/TLI values were greater than .95, RMSEA values were less than .08, and SRMR values were less than .08 (Hu & Bentler, 1999).
