Table 2.

Hierarchical regression model predicting concentrations of NAA. NAA model was used as a control metabolite to demonstrate the specificity of the model for choline. In the first and second stages GPC+PC concentration and the number of reversals are included as covariates of no interest. In the third stage reinforcement learning model parameter estimates, perseverative and regressive errors are included in the model.

Variable	Unstandardised β / Standardised β	t	R	R2 / Adj R2	ΔR2
Stage 1			.294	.086 /.003
GPC+PC	.802 /.294	1.018
Stage 2			.504	.254 /.105	.168
GPC+PC	.091 /.033	.103
Reversals	.117 /.485	1.501
Stage 3			.889	.790 /.497	.536
GPC+PC	1.836 /.672	1.628
Reversals	-.018 / − .073	-.181
${\mathrm{\alpha }}^{+}$	4.083 /.967	2.701 *
${\mathrm{\alpha }}^{-}$	-16.326 / − 1.092	-3.164 *
β	1.367 /.976	2.703 *
Perseverative err	.059 /.995	2.050
Regressive err	.045 /.668	1.417

N = 13; * p < .05; ** p < .01; *** p < .001