Table 3.

Changes in Fit of Bivariate Latent-Change-Score Models With the Addition of Coupling Parameters

Model	Parameters	χ2	df	CFI	BIC	w(BIC)	RMSEA	Δχ2	Δdf	p
Without covariates
No coupling	33	186	119	.990	34,194	.251	0.023
Fluid intelligence → Δdepressive symptoms	34	178	118	.991	34,192	.681	0.022	−8	1	.003
Depressive symptoms → Δfluid intelligence	34	185	118	.990	34,200	.012	0.023	−1	1	.282
Full coupling	35	176	117	.992	34,197	.056	0.021	−10	2	.005
With covariates
No coupling	77	415	347	.987	21,775	.066	0.019
Fluid intelligence → Δdepressive symptoms	78	404	346	.989	21,770	.802	0.018	−11	1	.001
Depressive symptoms → Δfluid intelligence	78	411	346	.988	21,777	.024	0.019	−4	1	.040
Full coupling	79	401	345	.989	21,774	.108	0.018	−14	2	.001

Note: Parameter = estimated model parameter; χ² = deviance (−2 × log-likelihood); CFI = comparative fit index; BIC = Bayesian information criterion; w(BIC) = Schwarz weight (i.e., relative probability of model preference); RMSEA = root-mean-square error of approximation; Δχ² = change in model misfit with addition of coupling (or couplings) compared with no-coupling model (lower values = better fit); p = p value for likelihood-ratio test of change in model fit. The RMSEA 95% confidence interval was within ±.007 for all models.