Figure 2: SampEn simulations using white and 1/f noise.
From: On the estimation of brain signal entropy from sparse neuroimaging data
(A) Accuracy and precision of SampEn (MSE at scale 1) estimates for white noise and 1/f noise as a function of signal length (number of successive data points). The average SampEn across 1,000 randomly generated time series with 216 (65,536) continuous data points served as the benchmark for accuracy estimation. (B) Accuracy and precision of SampEn estimates for white noise and 1/f noise as a function of number of pattern comparisons when estimating SampEn across discontinuous segments of 128, 32, and 3 data points. For better comparability, SampEn estimates from successive data points (with equating the number of pattern comparisons, see Methods) are also provided [color coding as in (A)]. Note. Error bars indicate the SD of the SampEn values across 1,000 randomly generated time series, hence represent the precision of the SampEn values. Accuracy and precision are given in log10 scaling (with smaller values plotted upwards) for better readability. SampEn = sample entropy; SD = standard deviation; Δ = absolute average difference between benchmark SampEn and SampEn estimated from shorter or discontinuous time series; discont. = discontinuous segments.
