. Author manuscript; available in PMC: 2021 Apr 1.

Published in final edited form as: Dev Psychol. 2016 May 5;52(6):943–959. doi: 10.1037/dev0000117

Table 2.

Regression coefficients and utility for sets of predictors.

	Regression coefficients Sample A		Regression coefficients Sample B		Utility Sample A	Utility Sample B
Variable	B	SE	B	SE	ΔR²	ΔR²
(Intercept)	−0.32^***	0.030	−0.32^***	0.032
Female	0.072^***	0.022	0.088^***	0.023	.0011^**	.0016^***
Socioeconomic status
Log-Income	−0.32^***	0.029	−0.28^***	0.028	.0185^***	.0129^***
Years of education	0.015^**	0.005	0.010^*	0.005	.0012^***	.0005^*
Work status					.0020^***	.0006
Working full-time	0.116^***	0.028	0.067^*	0.029
Other work status	0.075^*	0.032	0.034	0.034
Living arrangements					.0043^***	.0048^**
Household size
1 person	−0.110^*	0.054	−0.074	0.052
3 or more persons	0.097^**	0.040	0.084^*	0.039
Living with children	0.054	0.044	0.101^*	0.041
Relationship status					.0027^***	.0088^***
Not living with a partner	0.10^*	0.051	0.10	0.055
Single	0.20^***	0.051	0.33^***	0.046
Functional limitations					.0299^***	.0233^***
somewhat	0.22^***	0.026	0.21^***	0.028
severe	0.58^***	0.044	0.52^***	0.044
Social engagement					.0006	.0016^**
Volunteering	−0.008	0.010	−0.021	0.011
Political engagement	−0.032	0.023	0.012	0.023
Religious engagement	−0.014	0.012	−0.033^*	0.013
No. of friends	−0.028^***	0.003	−0.021^***	0.003	.0101^***	.0047^***
Contact frequency					.0214^***	.0198^***
Face-to-face contact with friends	−0.111^***	0.014	−0.118^***	0.015
Face-to-face contact with relatives	−0.051^***	0.012	−0.069^***	0.013
Contact with friends and relatives abroad	0.038^***	0.009	0.016	0.009
Use of social online networks	0.051^***	0.007	0.039^***	0.007

Notes. All continuous predictors were centered on the mean of the total sample. Standard errors are adjusted for clustering. Categorical predictors were dummy-coded with the following reference groups: gender: male; work status: not working; household size: 2 persons; relationship status: living with partner; functional limitations: not at all. The utility ΔR² corresponds to the difference in R² between the full model reported here and a reduced model that did not include the predictors for which the utility was computed.

^***

p < .001,

^**

p <. 01,

p < .05.