+
From a given vector a of measurements and a second
vector p with relative probabilities for each measure, we build the
probability distribution for the measured quantity. Then we output just
three quantities the mean and the 2-sigma low and high errors.
IN: a - fltarr measurements
p - fltarr relative prob. for each measurement
(normalization NOT required)
binsize - float binsize for grouping the measurements
OUT: m - float mean
low- float 2-sigma error on the left ()
high-float 2-sigma error on the right side
limit is the one-side Gaussian integrated probability to be reached
for 1-sigma limits p=0.68 and limit=0.34, for 3-sigma limits p=0.997 and
limit=0.499, for 2-sigma limits p=0.955 and limit=p/2.
in general n-sigma=gauss_cvf((1.-p)/2)
C. Allende Prieto, UT, Dec 2002