# Table 2 Age-specific direct and indirect relationship between depressive symptoms and cognitive functioning through BMI

Variables Mediator Outcome
BMI Cognitive Function
$$\beta$$ SE t p $$\beta$$ SE t P
CES-D × Age ≥ 70 -0.080** 0.026 -3.149 0.002
Age ≥ 70 -0.110 0.115 -0.956 0.339
CES-D 0.004 0.019 0.192 0.848 -0.323** 0.023 -14.352 0.000
BMI      0.139** 0.022 6.204 0.000
Covariates
Sexa1 -0.319** 0.077 -4.134 0.000 0.614** 0.129 4.758 0.000
Family Incomeb 0.125** 0.046 2.720 0.007 0.958** 0.072 13.253 0.000
ADL -0.139** 0.045 -3.099 0.002 -2.084** 0.075 -27.884 0.000
Marrieda2 0.094 0.092 1.015 0.310 1.431** 0.153 9.328 0.000
Meet friendsa3 -0.082 0.095 -0.866 0.387 1.777** 0.160 11.134 0.000
Good healtha4 -0.198** 0.086 -2.291 0.022 1.317** 0.142 9.254 0.000
Regular Exercisea5 0.151* 0.077 1.948 0.052 1.144** 0.130 8.804 0.000
R2 = 0.020, F (10, 5608) = 11.547, p < .000 R2 = 0.340, F (9, 5609) = 320.456, p < 0.000
Indirect relationship between Depressive Symptoms and Cognitive Function through BMI
Age Estimate SE 95% LLCI 95% ULCI
Age < 70 0.001 0.003 -0.005 0.006
Age ≥ 70 -0.011** 0.003 -0.018 -0.005
The Difference in Indirect relationship across Two Age Groups
Estimated difference SE 95% LLCI 95% ULCI
(Age ≥ 70)—(Age < 70) -0.0111** 0.0042 -0.0203 -0.0037
1. Abbreviations: SE Standard errors, t is t-statistics and p is the corresponding p-value
2. aDummy variable indicating group: 1 = Men; 2 = Currently married with a living spouse; 3 = Meet friends at least once a month; 4 = have a good health status; 5 = Exercise regularly
3. bLogarithm of family income is used in the estimation model
4. Standard errors are bootstrapped using 5,000 bootstrap replications
5. ** p < 0.05; * p < 0.10