Robin Ince

The previous post on quantization dealt with binning a continuous variable into a small number of discrete categories. But sometimes the source data is already discrete, for example when the data are counts (e.g. counts of spikes emitted by a neuron within a certain time window). Sometimes the raw...

In my work I make heavy use of mutual information. One interpretation of mutual information is as (with a sample size dependent scale factor) the effect size for a likelihood-ratio test of independence. This is similar to a Chi-squared test of independence; in fact the chi-squared test...

I found that some of the MEG data I am working with has long tails which was causing some problems when using non-robust statistics that assume normality. Visual inspection showed a t-distribution might be a good fit: The figures show the histogram of the data, together with the...