Fix for bug #12456 (https://gna.org/bugs/?12456).
When calculating the errors from replicate spectra, the standard deviations of all spins were being averaged rather than the variance. The same error occurred when averaging across spectra - when not all are duplicated, triplicated, etc.
Authored by: bugman 2008-10-16
Parent: [r7783]
Child: [r7785]
