Figure 2.

Structure of the main routines in the ToolBox. The core function is entropy.m which allows to compute the entropy-like quantities which constitute the building blocks for the calculation of mutual information and of all the other breakdown terms. It is important to note how the modular structure allows for new methods and extrapolation procedures to be easily created and linked to the existing software. In particular, starting with the next release of the toolbox, users will be given the opportunity to easily plug-in their custom bias correction m-files in order to extend the capabilities of the existing code. Other routines in the toolbox are available for the pre-processing of the input to entropy.m. buildr.m allows to build the response-matrix to be fed to the information routines starting from L + 1 -long arrays: the first array stores a list specifying, for each trial, which stimulus was presented to the subject while the L remaining arrays specify the L corresponding recorded responses. binr.m allows to discretize continuous response-matrices – prior to calls to entropy.m with the the Direct Method – according to a binning method chosen among a list of available binning options. Again, users are given the possibility to define their own binning strategy (in addition to the built-in binning methods) and to easily link their code to the toolbox. Finally, information.m is a wrapper around entropy.m which directly computes the breakdown terms by combining the outputs from this main function, thus skipping all the complex computations necessary for the the calculation of the breakdown quantities.