Table 5.

Multilevel models to test Hypotheses 3 and 4

Predictors	Episodic memory			Executive functioning			BTACT composite
Predictors	Estimates	95% CI	p	Estimates	95% CI	p	Estimates	95% CI	p
Intercept (M2)	−0.41	−0.59 to −0.23	<0.001	−0.05	−0.18 to 0.08	0.450	−0.11	−0.23 to 0.01	0.085
MIDUS wave	−0.13	−0.26 to 0.00	0.051	−0.43	−0.51 to −0.36	<0.001	−0.38	−0.45 to −0.31	<0.001
Exponent	0.26	−0.19 to 0.71	0.254	0.40	0.09 to 0.71	0.012	0.39	0.10 to 0.68	0.009
Sex	0.71	0.49 to 0.93	<0.001	0.08	−0.08 to 0.24	0.306	0.19	0.04 to 0.34	0.012
Education	0.17	0.04 to 0.30	0.008	0.13	0.04 to 0.21	0.003	0.13	0.05 to 0.21	0.002
Lag between waves	−0.08	−0.19 to 0.04	0.187	−0.19	−0.28 to −0.11	<0.001	−0.18	−0.26 to −0.10	<0.001
Wave × exponent	−0.02	−0.48 to 0.44	0.926	−0.21	−0.47 to 0.05	0.120	−0.16	−0.42 to 0.09	0.204
Random effects
σ²	0.47			0.15			0.14
τ₀₀	0.44 _M2ID			0.28 _M2ID			0.25 _M2ID
ICC	0.48			0.65			0.65
N	234 _M2ID			235 _M2ID			234 _M2ID
Observations	451			458			443
Marginal R²/Conditional R²	0.136/0.551			0.179/0.710			0.182/0.710

Repeated measured cognitive data nested within participant, indicated by M2ID. Sex coded as 0 = male, 1 = female. Education coded as −1 = high school or less, 0 = some college, 1 = bachelor's degree or higher. MIDUS wave coded as 0 = MIDUS 2, 1 = MIDUS 3. Years between waves (i.e., between MIDUS 2 and MIDUS 3 Cognitive Projects) and age at M2 Neuroscience Project are mean centered. Empirical Bayes slope estimation used (Bates et al., 2015).
Bold values are significant at p < .05.