Table 7.

Regression coefficients predicting log(k) as a function of DCM parameters.

Task period	DCM parameter	Elastic net parameter
Fixed	v −> d	0.6
	d −> v	−1.9
All choices	v −> d	−0.7
	v self	−0.1
	d −> v	3.3
	d self	−1.6
Later choices	v −> d	0.8
	v self	0.0
	d −> v	−0.9
	d self	0.2
Driving inputs	Value −> v	0.4
	Accept −> d	0.6

This table reports the regression coefficients from the elastic net regression using the DCM parameters specified above to predict discount rates (log(k)).

The regression is identical to those used in the prediction exercises described in the main text except that it was run with all 27 subjects at once. The regressions were estimated using the DCM parameters for both functional MRI runs separately and the values listed in this table represent the average of coefficients across runs for conciseness and clarity.

Note that the DCM parameters have been z-scored across subjects so that comparing the coefficients across parameters shows the relative size and direction of the effects of each DCM parameter on discount rates. See the main text of the Results section for quantification of the influence of each parameter on the model's ability to predict discount rates. Also, recall that smaller k values indicate lower discount rates and, therefore, parameters with negative regression coefficients increase the likelihood of choosing delayed reward options in our temporal discounting paradigm.

The labels Fixed, All Choices, Later Choices, and Driving Inputs correspond the portions of the DCM described in the main text and shown in Figure 5.

The label v −> d refers to signaling from vmPFC to dlPFC-BA 46.

The label d −> v refers to signaling from dlPFC-BA 46 to vmPFC.

The label v or d self refers to the parameters for the inhibitory self-connections to each region at the specified time points.

The label Value −> v refers to an input equal to rdSV from GLM-rdSV into vmPFC.

The label Accept −> d refers to an input equal the Accept regressor specifying later choices in GLM-rdSV.