def binning_prof (raw, knum=200, NBmin=100, plot=True):
'Do binning analysis.\
knum: number of bin-length to be simulated. (k=bin size)\
NBmin: minimum number of bins.\
Return: [uncorrelate_data(NBmin)], [bin_size(knum)], [auto_correlation_time(knum)]'
uncorr_data, ks, corrtime, err = [], [], [], []
# Calcualte kmax and kmin, bin length is graw as n*kmin, n=1,2,...
kmax = len(raw) / NBmin
if kmax == 0: kmax = 1
kmin = kmax / knum
if kmin == 0: kmin = 1
knum = kmax / kmin
# Mearge the bin with length k=kmin
basedata = _merge_bin (kmin, raw)
# Get "_ks" and "_corrtime". "_uncorr_data" remains with the largest bin-length.
var0 = st.var (raw)
for q in range(1, knum+1):
uncorr_data = _merge_bin (q, basedata)
k, var = q*kmin, st.var (uncorr_data)
err.append ([k,st.err(uncorr_data)])
ks.append (k)
corrtime.append (_auto_corr_time (k, var, var0))
if plot:
pl.plot (ks, corrtime, marker='.')
pl.xlabel ('bin size')
pl.ylabel ('auto correlation time')
pl.show()
return uncorr_data, ks, corrtime
def binning (raw, k): uncorr = _merge_bin (k, raw) err = st.err(uncorr) * 2 return st.mean (uncorr), err
