was discretized such that the values across trials were ranked and binned into three states (low, medium, or high firing) with equal number of counts (or as close to equal as possible in the event of tied values). The probability was then calculated by dividing the number of joint state observations by the total number of trials. For instance, in the case of drink outcome encoding, for a given neuron and time bin, we may have observed five joint states in which the neuron had a low firing rate during drinking trials. In this example, pi,t(x = low,y = drinking) = 5/20 = 0.25. In the case of stimulus encoding, we might have observed seven joint states in which the neuron had a high firing rate during real trials. In this example, pi,t(x = high,y = real) = 7/40 = 0.175. Note that drink encoding only used real trials, so only 20 total observations where performed, whereas trial encoding used both real and null trials, resulting in 40 total observations.
