new 'non parametric' discount function estimation

[drbenvincent]

Add the ability to estimate a 'non parametric' discount function. This will amount to an approach very similar to Frye et al (2016) in that we are attempting to locate indifference points at a number of delay 'slices'. It is different however in that it is using the Bayesian Adaptive Design approach. It would effectively be doing multiple Psychometric Function Estimation.

Step 1: implement a single DelaySlice model

This will basically be BAD applied to a psychometric function.

• get a single DelaySlice model implemented
• add some basic diagnostic plotting to check it's working as intended

Step 2: do 'non parametric' estimation by wrapping multiple DelaySlice models for different delays

We will use multiple instances of the DelaySlice model, each set up for a different delay. Each trial can then focus on one of the particular designs.

• create a notebook with a simple wrapper code to do this. Use the code in the multiple simultaneous model estimation notebook as a starting point for this.

Future work

• Implement it within PsychoPy so we can run experiments on people to do this.
• Extend to risky choice tasks.

[drbenvincent]

[drbenvincent]

[drbenvincent]