def do_the_fit(self):
options = theta_auto.Options()
options.set('minimizer', 'strategy', 'robust')
self.fit_res = theta_auto.mle(
self.model,
"data",
1,
chi2=True,
options=options,
with_covariance=True,
)[self.template_names[-1]]
print self.fit_res
par_values = {
"beta_signal": self.fit_res["beta_signal"][0][0],
}
for tmplt_name in self.template_names[:-1]:
bg_name = "bg_" + tmplt_name
par_values[bg_name] = self.fit_res[bg_name][0][0]
self.val_err = []
for tmplt_name in self.template_names[:-1]:
val = self.model.get_coeff("histo", tmplt_name).get_value(
par_values)
err = self.fit_res["bg_" + tmplt_name][0][1]
self.val_err.append((val, err))
self.val_err.append((
self.fit_res["beta_signal"][0][0],
abs(self.fit_res["beta_signal"][0][1])
))
def __init__(
self,
model_func,
input_path='../HistoLoader',
input_path_sys='../HistoLoaderSys',
filter_keyfunc=None,
hook_loaded_histos=None,
asymptotic=True,
limit_func=None,
cat_key=lambda _: 'histo', # lambda w: w.category,
dat_key=lambda w: w.is_data or w.is_pseudo_data,
sig_key=lambda w: w.is_signal,
bkg_key=lambda w: w.is_background,
sys_key=None,
tex_table_mod_func=tex_table_mod,
selection=None,
pvalue_func=None,
postfit_func=None,
name=None,
):
super(ThetaLimits, self).__init__(name)
self.model_func = model_func
self.input_path = input_path
self.input_path_sys = input_path_sys
self.filter_keyfunc = filter_keyfunc
self.hook_loaded_histos = hook_loaded_histos
self.cat_key = cat_key
self.dat_key = dat_key
self.sig_key = sig_key
self.bkg_key = bkg_key
self.sys_key = sys_key
self.tex_table_mod = tex_table_mod_func
self.model = None
self.what = 'all'
self.with_data = True
self.pvalue_func = pvalue_func# or (lambda m: theta_auto.zvalue_approx(
# m, input='data', n=1)
self.postfit_func = postfit_func or (lambda m: theta_auto.mle(
m, input='data', n=1))
self.selection = selection or self.name
self.mass_points = []
self.limit_func = limit_func or (
theta_auto.asymptotic_cls_limits
if asymptotic else
lambda m: theta_auto.bayesian_limits(m, what=self.what)
)
def __init__(
self,
model_func,
input_path='../HistoLoader',
input_path_sys='../HistoLoaderSys',
filter_keyfunc=None,
hook_loaded_histos=None,
asymptotic=True,
limit_func=None,
cat_key=lambda _: 'histo', # lambda w: w.category,
dat_key=lambda w: w.is_data or w.is_pseudo_data,
sig_key=lambda w: w.is_signal,
bkg_key=lambda w: w.is_background,
sys_key=None,
tex_table_mod_func=tex_table_mod,
selection=None,
pvalue_func=None,
postfit_func=None,
name=None,
):
super(ThetaLimits, self).__init__(name)
self.model_func = model_func
self.input_path = input_path
self.input_path_sys = input_path_sys
self.filter_keyfunc = filter_keyfunc
self.hook_loaded_histos = hook_loaded_histos
self.cat_key = cat_key
self.dat_key = dat_key
self.sig_key = sig_key
self.bkg_key = bkg_key
self.sys_key = sys_key
self.tex_table_mod = tex_table_mod_func
self.model = None
self.what = 'all'
self.with_data = True
self.pvalue_func = pvalue_func # or (lambda m: theta_auto.zvalue_approx(
# m, input='data', n=1)
self.postfit_func = postfit_func or (
lambda m: theta_auto.mle(m, input='data', n=1))
self.selection = selection or self.name
self.mass_points = []
self.limit_func = limit_func or (
theta_auto.asymptotic_cls_limits if asymptotic else
lambda m: theta_auto.bayesian_limits(m, what=self.what))
def do_the_fit(self):
self.fit_res = theta_auto.mle(self.model, "data", 1, chi2=True)
print self.fit_res
import os
import theta_auto
theta_auto.config.workdir = os.getcwd()
theta_auto.config.report = theta_auto.html_report(os.path.join(theta_auto.config.workdir, 'index.html'))
theta_auto.config.theta_dir = "/afs/cern.ch/work/h/htholen/private/cmsWorkingDir/CMSSW_5_3_11_patch6/theta"
model = theta_auto.build_model_from_rootfile("ThetaHistos.root", include_mc_uncertainties=True)
model.set_signal_processes(["real"])
model.add_lognormal_uncertainty("fake_rate", 1., "fake")
res = theta_auto.mle(model, "data", 1, chi2=True)
print res
par_values = {"beta_signal":res["real"]["beta_signal"][0][0], "fake_rate":res["real"]["fake_rate"][0][0]}
bkg_val = model.get_coeff("chhadiso", "fake").get_value(par_values)
bkg_err = bkg_val * res["real"]["fake_rate"][0][1] / res["real"]["fake_rate"][0][0]
sig_val = res["real"]["beta_signal"][0][0]
sig_err = res["real"]["beta_signal"][0][1]