New model: Hyperboloid discount function

[drbenvincent]

Add new model, Hyperboloid, where the discount fraction is 1/[(1+kD)^S] Note that this is equivalent to Exponential discounting of subjective time, according to Webber's Power Law.

Parameter: S

• We want to constrain S>-1, because below this value the curves accelerate.
• A reasonable approach is to set the mode of the prior equal to 1, as this equates to hyperbolic discounting. If the prior is reasonably broad then this will of course allow for deviations from 1.

Parameter: k

• We will treat the k parameter in exactly the same way as the hyperbolic models. This includes actually making inferences about log(k)

Models

We now have the following models available for use:

• ModelSeparateHyperboloid
• ModelMixedHyperboloid
• ModelHierarchicalHyperboloid

[drbenvincent]

This all seems to work very nicely now. Priors are acceptable, but as always could be refined given more empirical data.