Welch's method

Welch's method estimates the PSD by dividing the record into overlapping segments (typically 50 percent overlap), tapering and computing a periodogram for each, and averaging the periodograms. The averaging reduces variance in proportion to the number of independent segments; the overlap recovers some of the information lost at the tapered segment edges. It is the workhorse spectral estimator of applied BCI because it is simple, fast and robust.

The core tradeoff is set by the segment length: long segments give fine frequency resolution but few averages (high variance), while short segments give coarse resolution but many averages (low variance). Because each segment is short, Welch's method sacrifices low-frequency detail, since the lowest resolvable frequency is set by the segment length rather than by the whole record, which multitaper avoids. Overlapping segments are correlated, so the effective number of independent averages is smaller than the segment count.