Table 3.

Estimates of association between walnut consumption groups and trajectories of Telephone Interview for Cognitive Status scores from latent growth models

	Model 1						Model 2
	Intercept			Slope			Intercept			Slope
Walnut consumption	b		se	b		se	b		se	b	se
Low walnut consumption	1·53	***	0·21	0·01		0·03	0·73	***	0·17	–0·02	0·02
Moderate walnut consumption	2·22	***	0·27	–0·03		0·04	0·89	***	0·23	–0·03	0·03
Test effects
# tests 2012	0·37	***	0·04				0·14	***	0·04
# tests 2014	0·35	***	0·04				0·11	**	0·04
# tests 2016	0·30	***	0·04				0·05		0·04
Estimated mean intercept	17·84	***	0·42	–0·28	***	0·02	22·99	***	5·87	0·22	0·20
Intercept – slope correlation	0·14	**	0·05				0·15	***	0·04

Unconditional mean intercept b = 21·98, se = 0·09, P < 0·001; unconditional mean slope b = –0·26, se = 0·01, P = 0·01; unconditional correlation between intercept and slope r = 0·14, se = 0·06, P = 0·016.

Model 2 adjusted for the following covariates: gender, age, race/ethnicity, marital status, retirement status, education, longest occupational tenure, log-transformed household income, log-transformed household assets, BMI, vigorous physical activity, smoking status, alcohol consumption, number of doctor-diagnosed chronic conditions and number of mobility limitations.

***P < 0·001, **P < 0·01, *P < 0·05, †P < 0·10.