Table 4. Model Results for Sample 2: Unstandardised Path Coefficients, Random Slope Variances, and Indirect Effect Estimates.

Parameter	Self-focused attention				Negative social cognitions				Depressed mood
	Estimate	SE	p	PM	Estimate	SE	p	PM	Estimate	SE	p	PM
a	-0.016	0.001	< .001		-0.016	0.001	< .001		-0.036	0.005	< .001
b	0.796	0.250	.001		3.199	0.448	< .001		0.428	0.123	< .001
c’	-0.128	0.011	< .001		-0.092	0.010	< .001		-0.127	0.011	< .001
Vara	< 0.001	< 0.001	< .001		< 0.001	< 0.001	< .001		0.001	< 0.001	< .001
Varc’	0.007	0.001	< .001		0.005	0.001	< .001		0.007	0.001	< .001
Indirect effect ab	-0.013	0.004	.003	9	-0.050	0.008	< .001	35	-0.015	0.005	.002	11
Models reversing process variable and outcome
a	-0.158	0.012	< .001		-0.158	0.012	< .001		-0.159	0.012	< .001
b	0.030	0.007	< .001		0.040	0.005	< .001		0.078	0.021	< .001
c’	-0.010	0.002	< .001		-0.007	0.001	< .001		-0.021	0.005	< .001
Vara	0.010	0.002	< .001		0.010	0.002	< .001		0.010	0.002	< .001
Varc’	< 0.001	< 0.001	< .001		(see notes)				0.001	< 0.001	.001
Indirect effect ab	-0.005	0.001	< .001	32	-0.006	0.001	< .001	47	-0.012	0.004	.001	37

Note. n = 86. Path a represents the effect of time on the process variable. Path b represents the effect of the process variable on social anxiety score at the subsequent assessment (with time held constant). Path c’ represents the effect of time on social anxiety score controlling for the effect of the process variable. Path ab represents the indirect, or mediated, effect. The ‘reversed’ models swap the process and outcome variables. SE = standard error, P_M = percent mediation (i.e. the percentage of the total effect of time on outcome score that is accounted for by the mediated path ab), Var = variance. Due to lack of model convergence when the c’ path was specified as random, the reversed model for negative social cognitions was run using a fixed c’ path therefore no variance is given.