| REJKURT | Calculation of kutosis of a 1D, 2D or 3D array and rejection of outliers values of the input data array using the discrete kutosis of the values in that dimension. |
| Usage: | >> [kurtosis rej] = rejkurt( signal, threshold, kurtosis, normalize); |
| Inputs: | |
signal |
one dimensional column vector of data values, two dimensional column vector of values of size sweeps x frames or three dimensional array of size component x sweeps x frames. If three dimensional, all components are treated independantly. |
threshold |
Absolute threshold. If normalization is used then the threshold is expressed in standard deviation of the mean. 0 means no threshold. |
kurtosis |
pre-computed kurtosis (only perform thresholding). Default is the empty array []. |
normalize |
0 = do not not normalize kurtosis. 1 = normalize kurtosis. Default is 0. |
| Outputs: | |
kurtosis |
normalized joint probability of the single trials (same size as signal without the last dimension) |
rej |
rejected matrix (0 and 1, size: 1 x sweeps) |
| Remarks: | The exact values of kurtosis depend on the size of a time step and thus can not be considered as absolute. This function uses the kurtosis function from the statistival matlab toolbox. If the statistical toolbox is not installed, it uses the 'kurt' function of the ICA/EEG toolbox. |
| See also: | realproba |