Effect of data segmentation on coherence estimates of cardiac rhythms

Both fibrillatory and nonfibrillatory cardiac rhythms were analyzed for coherence spectra while varying the segmentation of finite records of data. Results indicate that although the magnitude and mean of the magnitude-squared coherence function do not change significantly for nonfibrillatory rhythms, these same parameters show a marked decrease for fibrillatory rhythms. These findings are explained in terms of the bias and variance of the coherence estimate.<
>