Table 6.

Summary of evaluation of the structural models: consistency, accuracy, predictive relevance, effect sizes and global fit.

C.: E.E.V.	DE1	E.V.	R2 Value	Q2 Value	ΔR2	f2	Gof
Model A							0.4900
T. → O.P. = c	Sig		0.3275	0.1768
Model B							0.6241
Endogenous Latent Variables:
ACAP			0.5987	0.3147
H2a = T. → ACAP = a1	Sig	0.5987
O.P. (mediated by ACAP)			0.5387	0.4417	0.2112	0.4578
H1 = T. → O.P. = c′	Nsig	0.0445a
H2b = ACAP → O.P. = b1	Sig	0.5832
Model C							0.6654
Endogenous Latent Variables:
ACAP			0.5681	0.3001	−0.0306	−0.0708
H2a = T. → ACAP = a1	Sig	0.5681
Innovation			0.4130	0.4133
H3a = T. → Innovation = a2	Nsig	0.0692b
H4a = ACAP → Innovation = a3	Sig	0.4822
O.P. (mediated by ACAP and Inn.)			0.7777	0.6487	0.2390	1.0751
H1 = T. → O.P. = c′	Nsig	0.0070
H2b = ACAP → O.P. = b1	Sig	0.2447
H3b = Inn. → O.P. = b2	Sig	0.5260
Improvement of Model C over Model A					0.4502	2.0252

C.: E.E.V., Comparison of Models: Effects on Endogenous Variables; DE, Direct Effects; E.V., Explained Variance; T., Training; O.P., Organizational Performance; Inn., Innovation

¹Results from Table 5: Sig. denotes a significant direct effect; Nsig. denotes a non-significant direct effect.

^aThis sum (0.0446 + 0.5832) is not equal to R² (0.5387), note that the contribution of training to the explained variance of O. Performance is negative, but very small. This commonly occurs when the sign of the zero-order correlation is the opposite of the sign of the path coefficient (Menard, 2009). In Model B, the path coefficient c′ is negative, but very small and non-significant direct effect.

^bThis sum (0.0692 + 0.4822) is not equal to R² (0.4130), note that the contribution of training to the explained variance of Innovation is negative, but very small. In Model C, the path coefficient a 2 is negative, but very small and non-significant direct effect.