Power spectral estimation with error bars

Power spectral estimation using multitaper, single or multiple traces and estimating error bars.

Description

This procedure calculates the averaged single-sided power spectrum with proper normalization, so that if the signal is given in volts, the value is truely the power. It uses multiple tapers to window the data, and thereafter performs a padding. The procedure is inspired by the procedure from the chronux.org software package.

The averaging is taken over both the trials and over multiple tapers (estimates of spectrum for a single trial). These different estimates of the spectrum is also use to calculate standard error of the estimates by boot strap technique.

• MATLAB release: MATLAB 7.7 (R2008b)

[Ella Podvalny]

There is a bug in SE computation, as px is normalized by df but pow_deleteone is not. You can solve it by:

pow_a = pow_a ./ df;
px = sum(pow_a, 2) / (size(pow_a, 2));

in place of:

px=sum(pow_a,2)/(size(pow_a,2)*df);