Table 3.

Comparison of models incorporating a potentially biasing variable x for subject 7.

Model	npars	−2LogL	AIC	BIC	χ2-increase (relative to)	df	p
y, η on x	52	1010	1114	1244
η on x	40	1044	1124	1224	34.250 (y, η on x)	12	0.001
y(a), η(a) on x	45	1034	1124	1237
η(a) on x	39	1049	1127	1225	15.061 (y(a), η(a) on x)	6	0.020

y(a) denotes the agreeableness marker items, and η(a) denotes the agreeableness factor. We allowed for direct effects of x on the latent variables but did not establish whether these were significant.

χ²-differences are reported for nested models.