Figure 6. Matthews correlation as a measure of replay identification performance as a function of the number of tetrodes included (up to 50 random selections per tetrode number).

(a) Matthews correlation coefficient, content accuracy, relative detection latency and non-burst detection rate for REST (left) and RUN2 (right). Dots represent individual tests, lines represent the median. (b) Cross-validated RUN1 decoding performance as a function of the number of tetrodes. Top: median in-arm decoding errors. Bottom: fraction arm-correct position estimates. (c) Scatter plot of the relation between median in-arm decoding error and replay detection performance during REST (top) and RUN2 (bottom). Black lines represent the centroid for each color-coded number of tetrodes in vivo.

Figure 6—figure supplement 1. Extended characterization of replay detection performance as a function of number of tetrodes for REST.

(a,b) and RUN2 (c,d). (a,c) Top: sensitivity and false-positive rate. Bottom: false discovery and false omission rates. Diagonal gray lines in scatter plots represent isolines for informedness (top) and markedness (bottom) measures. (b,d) Optimal thresholds with $thet{a}_{mua}$expressed in S.D. (top) or per-tetrode spiking rate (bottom).

Figure 6—figure supplement 2. Dependence of MUA and sharpness metrics on the number of tetrodes.

(a) MUA rate per tetrode in REST. Top: example distributions of MUA rate for time bins outside and inside candidate replay bursts (with or without identified replay content) for a subset of 4, 9 and 14 tetrodes. Bottom, left: mean MUA rate as a function of the number of tetrodes. Note that the mean rate remains constant, but the variance increases with fewer tetrodes. Bottom, right: Hellinger distance between MUA rate distributions as a function of the number of tetrodes. (b) Sharpness of posterior probability distributions in REST. Top: example distributions of sharpness values for time bins outside and inside candidate replay bursts (with or without identified replay content) for a subset of 4, 9 and 14 tetrodes. Note shift towards higher sharpness values when more tetrodes are included. Bottom, left: mean sharpness value as a function of the number of tetrodes. Bottom, right: Hellinger distance between sharpness distributions as a function of the number of tetrodes. (c) MUA rate per tetrode in RUN2. Panels as in (a). (d) Sharpness of posterior probability distributions in RUN2. Panels as in (b).