#m.setEntropyPrior()
#m.setFirstDerivativePrior()
# add uncertainties
uncList = [
] #'sjcalib1030', 'eup', 'ecup'] + ['lup'+str(x) for x in range(0, 10+1)] + ['cup'+str(x) for x in range(0, 3+1)] + ['bup'+str(x) for x in range(0, 3+1)] + ['ewkup']
for k in uncList:
print "Getting histograms for syst. ", k
struth, srecoWithFakes, sbkg, smig, seff, snrt = getHistograms(
"out_ttallhad_psrw_Syst.root", k, "mttAsymm")
m.addUncertainty(k, sbkg, smig.project('y'))
# add migration uncertainty
uncUnfList = [] #["me", "ps"]
for k in uncUnfList:
m.addUnfoldingUncertainty(k, mig[k], eff[k])
# try a curvature-based prior
# first choose alpha using only a MAP estimate
#m.setEntropyPrior()
#m.setCurvaturePrior(1.0)
m.setFirstDerivativePrior(1.0)
#m.setGaussianPrior()
#m.setConstrainArea(True)
m.run(data)
# does the same for the pseudo-data
alpha = {}
alphaChi2 = {}
bestAlphaBias = {}
unf_orig_mode = m.hunf_mode
del m
n = None
m = Unfolder(bkg["A"], mig["A"], eff["A"], truth["A"])
m.setUniformPrior()
#m.setGaussianPrior()
#m.setCurvaturePrior()
#if fb > 0:
# m.setFirstDerivativePrior(fb)
for k in uncUnfList:
# use uncertainty at reconstruction level, by using the nominal truth folded with the alternative response matrix
#m.addUncertainty(k, bkg["A"], np.dot(truth["A"].val, response[k].val))
# one can also use a non-linear term in the unfolding
m.addUnfoldingUncertainty(k, bkg["A"], mig[k], eff[k])
print("Response")
print(m.response.val)
print(m.response_unfsyst["B"].val)
m.run(data)
#m.graph("modelWithSysts.png")
#m.setAlpha(0)
m.sample(50000)
# plot marginal distributions
m.plotMarginal("plotMarginal.%s" % extension)
for i in uncUnfList:
#m.plotNPMarginal(i, "plotNPMarginal_%s.%s" % (i, extension))