Figure 8.

Binary word representation for a data set with patterns. (A) An example rastergram with 20 trials and two patterns (pattern 1 for the bottom 10 trials and pattern 2 for the top 10 trials). (B) In the corresponding binary representation, the presence of a spike during an event is indicated by a 1 (dark rectangle) and the absence by a 0. (C) The event model from which the data in panel A were obtained. The event occupation is a mixture. The event reliability for pattern 1 is indicated by the dark bars, whereas the event reliability of pattern 2 is indicated by the white-filled bars. The event model corresponding to the null hypothesis is the sum of both bars. Each binary word can be represented as a number; for four events, this number is between 1 and 16 (the binary value plus 1). Each binary word occurs with a certain probability, n_w, which is different for (D) the pattern model compared to (E) the null hypothesis model. (D, E) The theoretical probabilities are represented by the gray bars, and the empirical distribution corresponding to the data in panel B is indicated by the stem graphs. For example, the most frequent empirically occurring word is 1010, indicated by the tallest stem at the eleventh word.