def computeAnalyticError(p):
"""
@brief compute the analytic error for cross spectrum
"""
specDir = 'spectra/'
patchDir = 'patches'
freqs = p['frequencies']
powerOfL = p['powerOfL']
global nPatches
nDivs, nPatches = speckMisc.getPatchStats(patchDir,freqs[0])
print "Found %d patches with %d sub-season divisions in each"%(nPatches, nDivs)
if p['fileTag'] is None:
fileTag = ""
else:
fileTag = p['fileTag']
# Read in the single freq auto spectra first
# (C_b + N_b)^AA
lbin,clbin_aa,binWeight = numpy.loadtxt("%s/clBinAutoGlobalMean_%dX%d.dat"\
%(specDir,freqs[0],freqs[0]),unpack=True)
# (C_b + N_b)^BB
lbin,clbin_bb,binWeight = numpy.loadtxt("%s/clBinAutoGlobalMean_%dX%d.dat"\
%(specDir,freqs[-1],freqs[-1]),unpack=True)
# (C_b+N_b)^{AB}
lbin,clbin_ab,binWeight = numpy.loadtxt("%s/clBinAutoGlobalMean_%dX%d.dat"\
%(specDir,freqs[0],freqs[-1]),unpack=True)
# Cross spectrum
lbin,clbin_cross,binWeight = numpy.loadtxt("%s/clBinCrossGlobalWeightedMean_%dX%d.dat"\
%(specDir,freqs[0],freqs[-1]),unpack=True)
print clbin_cross
clbin_aa /= lbin**powerOfL
clbin_bb /= lbin**powerOfL
clbin_ab /= lbin**powerOfL
clbin_cross /= lbin**powerOfL
# Uncertainty in a patch
# noise spectra
N_b_aa = (clbin_aa - clbin_cross)
N_b_bb = (clbin_bb - clbin_cross)
N_b_ab = (clbin_ab - clbin_cross)
print "N_b_ab",N_b_ab
print "N_b_aa",N_b_aa
print "N_b_bb",N_b_bb
if len(freqs) == 1:
g = open("%s/Nlbin_season_%sx%s.dat"%(specDir,freqs[0],freqs[0]),"w")
for i in xrange(len(lbin)):
g.write("%d %e\n"%(lbin[i],N_b_aa[i]/nDivs))
g.close()
C_b = clbin_cross
fac = 2.0
apoCorr = 1.
# For auto spectra, implement the apodization correction
if freqs[0]== freqs[-1]: apoCorr = apoCorrection(specDir,patchDir,p)
print "Apodization correction = %s"%apoCorr
poissonPiece = 0.
# Add in the poisson piece (for 148 only now)
if freqs[0] == freqs[-1] and freqs[0] == 148 and p['poissonAmps_148'] != None:
poissonPiece = poissonError(specDir,patchDir,freqs[0],p)
print "Poisson Estimate = %s"%poissonPiece
if nDivs == 1:
Theta = (clbin_aa*clbin_bb)/(fac*binWeight)
else:
Theta = (2*C_b**2 + C_b*(N_b_aa + N_b_bb)/nDivs + \
2./nDivs*C_b*N_b_ab \
+ (N_b_aa*N_b_bb + N_b_ab*N_b_ab)/(nDivs*(nDivs-1)))\
/(fac*binWeight) + poissonPiece
C_b_err = numpy.sqrt(Theta/nPatches)*apoCorr
g = open('%s/spectrum_%dx%d_%s.dat'%(specDir,freqs[0],freqs[-1], fileTag),"w")
for k in xrange(len(lbin)):
g.write("%d %e %e\n"%(lbin[k],lbin[k]**powerOfL*C_b[k],lbin[k]**powerOfL*C_b_err[k]))
g.close()
return 0 # success