[ENH] Expected values and expected utility

[bclenet]

Goal

Create a common method for EV / EU computation

Why this is useful

At least 6 teams explicitly used expected values / expected utility as regressors of their first level models.

• L1A8 : EV + EU

For the 1st level fMRI, we used the event-related design with 3 parametric modulators, gain loss, and Euclidian distance, following the design of Tom et al., (2007) in a sequential manner.

• 1P0Y : EV

The first regressor modelled trial decision responses, either 'accept' (coded as '1s') and 'reject' (coded as '-1s') responses and the second parametric modulation regressor modelled 'expected value' (as defined in Canessa et al., 2013).

• 43FJ : EV

Expected value onsets, duration 4s, PM expected value calculated in accordance with Canessa et al. 2013 J Neurosci. Design was based on Canessa et al. 2013 J Neurosci.

• 0JO0 : ?
  See general comments about EV

• E3B6 : EV

We used an approach similar to Tom et al 2007. All trials were modeled using a single condition (with a duration of 4 secs) and three additional parametric regressors were included : \na) param_loss: a regressor modulated by the loss vallue associated to that trial\nb) param_gain: regressor modulated by the gain vallue associated to that trial\nc) the Euclidean distance of the gain/loss combination from the indifference line (i.e. assuming lambda=2 and a linear value function).

• 9U7M : EV

To assess the neural effects of gain, two parametric regressors of decision phase activation were included corresponding to the potential gain amount associated with each decision in addition to the expected value or potential average payoff associated with each decision.

Timeline

No response

More specifically

This is related to PR #74 that must be reviewed

Open questions

No response