Table 3.

Results of the confirmatory factor analyses for UFS-S and UFS-P and common goodness of fit threshold criteria

		M1: Higher-order model	M2: Two-factor model	Model fit thresholds (Hu & Bentler, 1999; Schermelleh-Engel et al., 2003)
UFS-S N = 248	χ2/df	2.746	3.064	2–3 good, 3–5 permissible
	CFI	0.893	0.872	> 0.95 good, > 0.90 permissible
	TLI	0.883	0.861	> 0.95 good, > 0.90 permissible
	RMSEA	0.084	0.091	< 0.05 good, < 0.05-0.10 permissible
	SRMR	0.049	0.050	< 0.05 good, < 0.05-0.10 permissible
	AIC	14711.378	14825.334***	comparative index
	BIC	15024.073	15123.976***	comparative index
UFS-P N = 243	χ2/df	3.087	3.746	2–3 good, 3–5 permissible
	CFI	0.849	0.798	> 0.95 good, > 0.90 permissible
	TLI	0.828	0.774	> 0.95 good, > 0.90 permissible
	RMSEA	0.093	0.106	< 0.05 good, < 0.05-0.10 permissible
	SRMR	0.063	0.072	< 0.05 good, < 0.05-0.10 permissible
	AIC	12159.421	12286.566***	comparative index
	BIC	12393.457	12510.122***	comparative index

Notes UFS-S = UPSIDES fidelity scale service user version, UFS-P = UPSIDES fidelity scale peer support worker version, M1 = model with two factors, factor implementation as factor of higher order, divided into sub-categories. Sub-categories in UFS-S: receipt, engagement, enactment, competence. Sub-categories in UFS-P: receipt, competence, communication. M2 = model with two factors (implementation and active ingredients), no sub-categories. χ2 = Chi-square index, df = Degrees of freedom, CFI = Comparative fit index, TLI = Tucker-Lewis index, RMSEA = Root mean square error of approximation, SRMR = Standardised root mean square residual. AIC = Aikaike information criterion, BIC = Bayesian information criterion. *** Comparison of AIC and BIC Model 1 vs. Model 2: χ2 difference test p < .001