Table 2. Model Selection by Various Statistical Methods.

Method	Neuroticism	Extraversion	Openness	Agreeableness	Conscientiousness
Stepwise	NAA4 β = .27	CRE4 β = −.31		CRE4 β = −.43	CHO3 β = −.38
α = 0.05				NAA5 β = −.27	CRE5 β  =  .35
	r2 = .31	r2 = .12		r2 = .35	r2 = .25
Interaction				ns, β = −1.58	ns, β = −.57
Backward	CHO5 β = −.31	CRE2 β = .49		CRE4 β = −.43	CHO3 β = −.38
α = 0.05	CHO6 β = .34	CHO2 β = −.61		NAA5 β = −.27	CRE5 β = .35
		CHO5 β = .36
	r2 = .36	r2 = .17		r2 = .35	r2 = .25
Interaction	ns, β = −.24	ns, β = .02/.19/−.56		ns, β = −1.58	ns, β = −.57
AICc 1	NAA4	NAA2	NAA1	NAA5	CHO3
	CHO5	CRE2	NAA4	CRE2	CRE5
	CHO6	CHO2	CHO3	CRE4	(205.08)
	(210.21)	CHO5	(197.37)	(173.80)
		(220.96)
AICc 2	CHO5	NAA2	NAA1	CRE4	CHO3
	CHO6	CHO2	NAA4	CRE5	CRE5
	(201.51)	CHO5	CHO2	(174.01)	CHO2
		(221.47)	(197.51)		(206.50)
AICc 3	NAA4	NAA2	CHO2	NAA5	NAA2
	CHO4	NAA5	(197.66)	CRE2	CHO3
	CHO5	CRE4		CRE4	CRE5
	CHO6	CHO2		CHO2	(206.94)
	(212.01)	CHO5		CHO3
		(222.07)		(174.08)

Models selected by various statistical methods. Columns are individual response variables. All models include age, gender and handedness as covariates. Cell entries are variables selected from 18 metabolite-region combinations. If a cell is empty then no other variables are added. AICC 1, 2 and 3 are the three best models selected by the adjusted Akaike Information Criterion (AICC), which is widely viewed as the best model selection criteria available. This measure is penalized version of minus twice the maximized log-likelihood, where the penalty is a function of the number of predictors in the model. The value of AICC is given in parentheses.