R. Siedenberg et al., THE CORRELATION-COEFFICIENT AND THE GOODNESS-OF-FIT IN-SOURCE LOCALIZATION OF NOISE RECORDED BY MAGNETOENCEPHALOGRAPHY, Brain topography, 9(2), 1996, pp. 95-100
We recorded magnetic field activity to evaluate the specificity of the
correlation coefficient and the goodness of fit in source localizatio
n. Magnetic noise was recorded (bandpass 1-200 Hz) with a 37-channel m
agnetometer with and without a subject present in a magnetically shiel
ded room After averaging of the recorded epochs, the data were subject
ed to a single equivalent current dipole algorithm. The correlation co
efficient and the goodnesses of fit were computed for each dipole fit.
The number of high correlations and high goodnesses of fit was determ
ined. In recordings without a subject present the averaged original da
ta never achieved a correlation higher than 0.70 and a goodness of fit
higher than 0.85. After more restrictive filtering (bandpass 1-20 Hz)
there was a significant increase for both correlation and goodness of
fit. In recordings with a subject present (but not receiving external
stimulation) the correlation and the goodness of fit for the averaged
original data reached 0.96 and 0.97; after filtering, both measures r
eached values greater than or equal to 0.99. The results of this study
indicate that both the correlation coefficient and the goodness of fi
t lack the specificity to distinguish stimulus-related brain activity
from background noise.