Diagnostics display help

• Autocorrelation plot

What is it ?

It describes how much your data are auto-correlated.

What do you want ?

A beautiful center pick. (It is often more noisy)

What should alarm you ?

Several picks in the center (your signal is probably not aligned on your inputs defined by the model)

--> Something is missing in your model or your onsets are ill-defined.

A weird gaussian : sign of underfit (check your priors, they are probably too strong or there is a dynamic in your data your model is not taking into account).

• Volterra Kernels

What is it ?

Typical reaction of the model to the input. When you have lot of inputs, responses tend to overlap. Volterra Kernels allow you to un-overlap your responses and to obtain a typical response to your input. It can be useful for exploratory analysis : it can give you trial by trial the best estimate of some parameters (example of alpha in reversal learning) (you can see reaction that gives you an insight of how you could adjust your model)