Replay of RUN-order sequences.

During behavior place cells fire in sequences that are replayed during sleep. The fitted VLMC models encode the sequential correlations between cells during sleep and predict the probabilities of the RUN sequences and their subsequences (RUN-order sequences). The majority of these likelihood ratios indicate that they are enriched relative to avalanche independent model. However, the raw likelihood ratios do not directly attest to the statistical significance of this enrichment. We assessed significance using the z-score of the likelihood ratio computed using the values from 10-fold cross validation. These z-scores incorporate the ‘jitter’ in the model fitting and provide a strong signature of statistical significance of the likelihood ratio. The significance level was set at p < 0.05 after Bonferroni correction for multiple hypothesis testing. However, we note that the likelihood ratios for all sequences are correlated, so this significance level is stringent. The full RUN sequence was significantly enriched in 3 of 4 sessions (representing all three rats).