Table 3.

Mediation effects of changes in psychosocial factors on changes in fat intake

Psychosocial factor	Single-mediator models
	α(SE)	β(SE)	αβ(SE)	95% CI of αβ	z	% mediated effect
Attitude	-0.004 (0.031)	-7.773 (2.077)***	0.031 (0.241)	-0.441, 0.503	0.129	statistical suppressiona
Self-efficacy	-0.020 (0.077)	-1.790 (1.581)	0.036 (0.139)	-0.236, 0.308	0.258	statistical suppressiona
Social support	0.005 (0.098)	-2.548 (1.732)	-0.013 (0.250)	-0.502, 0.477	-0.051	0.1
Perceived benefits	0.069 (0.039)	-1.799 (2.997)	-0.124 (0.113)	-0.345, 0.097	-1.099	1.4
Perceived barriers	0.067 (0.022)**	2.260 (1.493)	0.151 (0.060)*	0.034, 0.268	2.535	suppression
	Multiple-mediator model
	α(SE)	β(SE)	αβ(SE)	95% CI of αβ	z	% mediated effect
Attitude	-0.004 (0.031)	-6.966 (1.866)***	0.028 (0.054)	-0.078, 0.134	0.515	statistical suppressiona
Self-efficacy	-0.020 (0.077)	-0.379 (1.941)	0.008 (0.040)	-0.070, 0.085	0.191	statistical suppressiona
Social support	0.005 (0.098)	-1.643 (2.177)	-0.008 (0.037)	-0.080, 0.060	-0.223	0.1
Perceived benefits	0.069 (0.039)	-0.144 (3.055)	-0.010 (0.211)	-0.423, 0.403	-0.047	0.1
Perceived barriers	0.067 (0.022)**	0.786 (1.741)	0.053 (0.117)	-0.177, 0.282	0.450	statistical suppressiona

α estimate of intervention effect (unstandardized regression coefficient) on residualized change score of psychosocial factors

β estimate of the independent effect of the mediator (unstandardized regression coefficient of residualized change scores) on residualized change score for fat intake αβ product-of-coefficient estimate; mediated effect

SE standard error

95% CI of αβ 95% confidence interval of the mediated effect

z standard deviate associated with mediated effect (used for significance testing)

* significant at the 5% probability level, ** significant at the 1% probability level; *** significant at the 0.1% probability level

a Statistical suppression refers to a statistically non-significant indirect effect that is opposite in sign to the intervention effect