def get_mc_fit( sampMan, sampname, sel_base, eta_cuts, xvar, plot_var, binning, workspace, suffix='' ) :
ph_selection_sr = 'ph_n == 1'
xmin = xvar.getMin()
xmax = xvar.getMax()
addtl_cuts_sr = 'ph_pt[0] > 50 && %s > %d && %s < %d ' %(plot_var, xmin, plot_var , xmax )
results = {}
for ieta in eta_cuts :
eta_str_sr = 'ph_Is%s[0]' %( ieta )
full_sel_sr = ' && '.join( [sel_base, ph_selection_sr, eta_str_sr, addtl_cuts_sr] )
hist_sr = clone_sample_and_draw( sampMan, sampname, plot_var, full_sel_sr , binning )
label = '%s_%s_%s'%(sampname, suffix, ieta)
fitManager = FitManager( 'dijet', 2, hist_sr, plot_var, ieta, xvar, label, options.useRooFit)
fit_distribution( fitManager, sampMan, workspace, logy=True )
results[ieta] = save_distribution( fitManager, sampMan, workspace, logy=True )
return results
def makevariableplots(samp,ptlist,fitrange,basesel="1",tag="",var="ph_pt[0]"):
parmnames = FitManager.ParamDCB
parmvals = {name:[] for name in parmnames}
parmerrs = {name:[] for name in parmnames}
parmufloats = {name:[] for name in parmnames}
xpoints, xerrs = [],[]
fm = FitManager("dcbp", xvardata = (0,200,"GeV"))
for ptrange in zip(ptlist[:-1],ptlist[1:]):
xpoints.append((ptrange[1]+ptrange[0])/2)
xerrs .append((ptrange[1]-ptrange[0])/2)
### NOTE NOTE fitting step NOTE NOTE
parmdict = fitting(samp,fm,ptrange,fitrange.get(ptrange[0],fitrange[-1]),basesel=basesel,tag=tag,var=var)
for name,val in parmdict.items():
parmvals[name].append(val.n) #value
parmerrs[name].append(val.s) #uncertainty
parmufloats[name].append(val)
# dealing with width of overflow/last bin
xpoints[-1] = ptlist[-2] + xerrs[-2]
xerrs[-1] = xerrs[-2]
print "values:"
pprint(parmufloats)
print "xpoints: \n", xpoints
print "xerrors: \n", xerrs
npt = len(ptlist)-1
tod = lambda x: array("d",x)
xpoints = tod(xpoints)
xerrs = tod(xerrs )
#for i in range(npt):
parmgraph = {name:ROOT.TGraphErrors(npt,
xpoints,tod(parmvals[name]),xerrs,tod(parmerrs[name])) \
for name in parmnames}
c1 = ROOT.TCanvas("c1","",800,500)
parmlimits = FitManager.setuparray["dcb"]
parmlimits ={n[0]:n[1:] for n in parmlimits}
for name, tgraph in parmgraph.items():
tgraph.SetTitle(name)
tgraph.GetXaxis().SetTitle("pt")
tgraph.GetYaxis().SetTitle(name)
yplot = parmrange[name]
tgraph.SetMaximum(yplot[1])
tgraph.SetMinimum(yplot[0])
setlinestyle(tgraph)
tgraph.Draw('AC*')
ylimit = parmlimits[name]
if len(ylimit)==3:
if (ylimit[2]<yplot[1]):
hline = addline(tgraph,ylimit[2])
hline.Draw()
print name," ", ylimit[2]
if (ylimit[1]>yplot[0]):
lline = addline(tgraph,ylimit[1])
lline.Draw()
print name," ", ylimit[1]
c1.SaveAs(options.outputDir+"/tge_dcb_%s_%s.pdf" %(name,tag))
c1.SaveAs(options.outputDir+"/tge_dcb_%s_%s.png" %(name,tag))
data = {'ptlist':ptlist,'parm':parmvals,'error':parmerrs}
with open('data/temp/dcbparms_%s.txt' %tag,'w') as outfile:
json.dump(data,outfile)
def makevariableplots_simultaneous(samp,ptlist,fitrange,basesel="1",tag="",ic = None, dobkgd=True, donorm=True, maxtimes=5, doData=False, var = "ph_pt[0]"):
""" extended pdf fit """
if ic is None: ic = dict(sig="dcbp",bkgd="gaus", ext="simul")
if "icparm" not in ic: ic["icparm"] = dict()
parmufloats, parmvals, parmerrs = defaultdict(list),defaultdict(list),defaultdict(list)
parmnames = FitManager.ParamDCB
parmnames2 = ["gauszg_mean","gauszg_sig"]
parm ,parm2= get_param("data/dcbparms_all.txt",ptlist), get_param("data/parms_zgamma.txt",ptlist)
parmptlist = json.loads(open("data/dcbparms_all.txt").read())['ptlist']
parm2ptlist = json.loads(open("data/parms_zgamma.txt").read())['ptlist']
fm = FitManager("dcbexpo", xvardata = (0,200,"GeV"))
## FIXME: need better ways to obtain fitted parameters
iconddcb = {pt:[(n,parm[n][parmptlist.index(pt)] ) for n in parmnames] for pt in ptlist[:-1]}
icondgzg = {pt:[(n,parm2[n][parm2ptlist.index(pt)]) for n in parmnames2] for pt in ptlist[:-1]}
#ieighty = parmptlist.index(80) # search for pt80 parameter
#iconddcb = {pt:[(n,parm[n][ieighty]) for n in parmnames] for ipt,pt in enumerate(ptlist[:-1])}
#icondgzg = {pt:[(n,parm2[n][ieighty]) for n in parmnames2] for ipt,pt in enumerate(ptlist[:-1])}
print "iconddcb"; pprint (iconddcb)
print "icondgzg"; pprint (icondgzg)
for ptrange in zip(ptlist[:-1],ptlist[1:]):
ptrange = tuple(ptrange)
ptname = tuple(map(lambda x: str(x).replace('.','p'), ptrange))
#xp.append(ptrange)
ic['icparm']['dcbp'] = [("x",20,200)]+iconddcb[ptrange[0]] # FIXME: move up a fx level
ic['icparm']['gauszg'] = [("x",20,200)]+icondgzg[ptrange[0]] # FIXME: move up a fx level
### fitting step
ptfitrange = fitrange.get(ptrange[0],fitrange[-1]) # get default if entry doesnt exist
values, stackcount = fitting_simultaneous(samp, fm, ptrange, ptfitrange, basesel=basesel, var = var,
tag=tag, ic = ic, xbins=(200,0,200), dobkgd =dobkgd, maxtimes = maxtimes, doData=doData, donorm=donorm)
if donorm: make_normalization_comparison(fm,values,stackcount,ptname,tag)
for name,val in values.items():
if val is None: #for case of non-initiated variable
parmvals[name].append(-1)
parmerrs[name].append(-1)
elif val.s>0: #check if the value is fitted
parmvals[name].append(val.n) #value
parmerrs[name].append(val.s) #uncertainty
parmufloats[name].append(val)
# dealing with width of overflow/last bin
#xpoints, xerrs = xp.output()
print "values:"; pprint(parmufloats.items())
f = lambda p: {x:y for x,y in p.iteritems()}
data = {'ptlist':ptlist,'parm':f(parmvals),'error':f(parmerrs)}
with open('data/parms_%s.txt' %tag,'w') as outfile:
json.dump(data,outfile)
def makevariableplots_simultaneous_zg( samp, ptlist, fitrange, basesel="1",
tag="", ic = None, dobkgd=True,
donorm=False, maxtimes=5, doData=False):
"""
extended pdf fit for zgamma peak
"""
if ic is None: ic = dict(sig="dcbp",bkgd="gaus", ext="simul")
if "icparm" not in ic: ic["icparm"] = dict()
parmufloats = defaultdict(list)
parmvals = defaultdict(list)
parmerrs = defaultdict(list)
parmnames = FitManager.ParamDCB
parm = get_param("data/efake/%i/dcbparms_all%i.txt"%((options.year,)*2))
fm = FitManager("dcbexpo", xvardata = (0,200,"GeV"))
iconddcb = {pt:[(n,parm[n][ipt]) for n in parmnames] \
for ipt,pt in enumerate(ptlist[:-1])}
for ptrange in zip(ptlist[:-1],ptlist[1:]):
ptrange = tuple(ptrange)
#xp.append(ptrange)
ic['icparm']['dcbp'] = [("x",50,200)]+iconddcb[ptrange[0]] # FIXME: move up a fx level
#ic['icparm']['gauszg'] = [("x",20,200)]+icondgzg[ptrange[0]] # FIXME: move up a fx level
### fitting step
ptfitrange = fitrange.get(ptrange[0],fitrange[-1]) # get default if entry doesnt exist
values, stackcount = fitting_simultaneous(samp, fm, ptrange, ptfitrange, basesel=basesel,
tag=tag, ic = ic, xbins=(200,0,200), dobkgd =dobkgd, maxtimes = maxtimes, doData=doData, donorm=donorm)
if donorm: make_normalization_comparison(fm,values,stackcount,ptrange,tag)
for name,val in values.items():
if val is None: #for case of non-initiated variable
parmvals[name].append(-1)
parmerrs[name].append(-1)
elif val.s>0: #check if the value is fitted
parmvals[name].append(val.n) #value
parmerrs[name].append(val.s) #uncertainty
parmufloats[name].append(val)
# dealing with width of overflow/last bin
#xpoints, xerrs = xp.output()
print "values:"; pprint(parmufloats.items())
f = lambda p: {x:y for x,y in p.iteritems()}
data = {'ptlist':ptlist,'parm':f(parmvals),'error':f(parmerrs)}
with open('data/efake/%i/parms_%s.txt' %(options.year,tag),'w') as outfile:
json.dump(data,outfile)
def fit_sample( hist, xvar, workspace , suffix, sample_params, label_config ):
fitManager = FitManager( 'dscb', hist=hist, xvardata = xvar,
sample_params=sample_params, label = suffix)
fitManager.setup_fit()
#fitManager.fit_histogram(workspace )
#fitManager.run_fit_minuit( fitrange = (fit_min, fit_max) ,debug = True)
fitManager.run_fit_minuit( fitrange = xvar, debug = True)
fitManager.get_results( workspace )
#fitManager.save_fit( sampMan, workspace, stats_pos='left',
# extra_label = extra_label , plotParam =True)
canv = fitManager.draw( subplot = "pull", paramlayout = (0.12,0.5,0.78),
useOldsetup = True, label_config = label_config)
canv.Print("%s/%s.pdf" %(options.outputDir, suffix) )
canv.Print("%s/%s.png" %(options.outputDir, suffix) )
canv.Print("%s/%s.C" %(options.outputDir, suffix) )
print "************"
print " RooFitResult Status: %d"%fitManager.roofitresult.status()
print "************"
fitmlist.append(fitManager)
def get_mc_fit(hist, var, fitfunc, workspace, suffix=''):
if 'mu' in suffix:
ch, extra_label = "mu", "Muon Channel"
elif 'el' in suffix:
ch, extra_label = "el", "Electron Channel"
xvar = ROOT.RooRealVar(var, var, _XMIN_M, _XMAX_M)
#full_sel_sr = sel_base.replace("ADDITION", " (%s > %d && %s < %d )"%(var, xmin, var , xmax))
#print full_sel_sr
#label = '%s_%s_%s'%(protag, suffix, fitfunc)
## power, expow, vvdijet, atlas, dijet
norders = 2
if fitfunc == "expow" or fitfunc == "atlas" or fitfunc == "vvdijet":
norders = 1
fitManager = FitManager(fitfunc, hist, xvar, suffix, norders=norders)
#fitManager.addhist(hist, "datahist"+fitfunc)
canv = fitManager.draw(paramlayout=(0.7, 0.5, 0.82),
useOldsetup=True,
logy=1,
yrange=(5e-3, 2e4))
#fitManager.setup_fit()
fitManager.setup_rootfit(xvar)
fitManager.func.Draw("same")
savecanv(canv, "%sbefore_%s" % (suffix, fitfunc))
fitManager.run_rootfit()
#fit_distribution( fitManager, sampMan, workspace, logy=True )
#fitManager.make_func_pdf()
#fitManager.fit_histogram( workspace )
print """ cast it from TF1 to RooGenericPdf """
fitManager.calculate_func_pdf()
fitManager.get_results(workspace, False)
outhist = fitManager.get_pdf_histogram(xbins=hist.GetNbinsX())
#results[ieta] = save_distribution( fitManager, sampMan, workspace, logy=True )
#fitManager.save_fit( sampMan, workspace, logy = True, stats_pos='right', extra_label = extra_label)
canv = fitManager.draw(subplot="pull",
paramlayout=(0.7, 0.5, 0.82),
useOldsetup=True,
logy=1,
yrange=(5e-3, 2e4))
savecanv(canv, "%s_%s" % (suffix, fitfunc))
return outhist
def make_signal_fits(sampMan,
sel_base,
eta_cuts,
plot_var,
xvar,
binning,
workspaces_to_save,
suffix,
plots_dir='plots'):
sampMan.clear_hists()
if not os.path.isdir(plots_dir):
os.makedirs(plots_dir)
for samp in sampMan.get_samples(isSignal=True):
print 'Sample = ', samp.name
res = re.match('(MadGraph|Pythia)ResonanceMass(\d+)_.*', samp.name)
if res is None:
print 'Could not interpret path ', samp.name
else:
mass = float(res.group(2))
iwidth = 0
if samp.name.count('width0p01'):
width = 0.0001
wid = "0p01"
else:
res2 = re.match(
'(MadGraph|Pythia)ResonanceMass(\d+)_width(\d)', samp.name)
width = float(res2.group(3)) / 100.
wid = res2.group(3)
iwidth = 1
if samp.name.count('MadGraph') == 0:
continue
if mass > 2000 or mass < 300:
continue
if iwidth >= len(binning):
print "exclude", iwidth, binning
continue
if mass not in binning[iwidth]:
print "exclude mass", mass, binning[iwidth]
continue
#ph_selection_sr = '%s==1' %defs.get_phid_selection('all')
#ph_idx_sr = defs.get_phid_idx( 'all' )
#print binning[iwidth][mass]
#addtl_cuts_sr = 'ph_pt[%s] > 50 && %s > %f && %s < %f ' %(ph_idx_sr, plot_var, binning[iwidth][mass][1], plot_var, binning[iwidth][mass][2] )
#xvar.setBins(10000,'cache')
#xvar.setMin('cache',-100)
#xvar.setMax('cache',1500)
weight_str = defs.get_weight_str()
sel_base_mu = defs.get_base_selection('mu')
sel_base_el = defs.get_base_selection('el')
el_ip_str = '( fabs( el_d0[0] ) < 0.05 && fabs( el_dz[0] ) < 0.10 && fabs( el_sc_eta[0] )<= 1.479 ) || ( fabs( el_d0[0] ) < 0.10 && fabs( el_dz[0] ) < 0.20 && fabs( el_sc_eta[0] )> 1.479 )'
el_tight = ' el_passVIDTight[0] == 1'
el_eta = ' fabs( el_eta[0] ) < 2.1 '
ph_str = 'ph_n==1 && ph_IsEB[0] && ph_pt[0] > 50 && !ph_hasPixSeed[0]'
ph_tightpt_str = 'ph_n==1 && ph_IsEB[0] && ph_pt[0] > 80 && !ph_hasPixSeed[0]'
met_str = 'met_pt > 25'
Zveto_str = 'fabs(m_lep_ph-91)>15.0'
sel_mu_nominal = '%s * ( %s && %s && %s )' % (weight_str, sel_base_mu,
ph_str, met_str)
sel_el_nominal = '%s * ( %s && %s && %s && %s && %s && %s && ( %s ))' % (
weight_str, sel_base_el, el_tight, el_eta, ph_str, met_str,
Zveto_str, el_ip_str)
sel_mu_phpt_nominal = '%s * ( %s && %s && %s && ADDITION)' % (
weight_str, sel_base_mu, ph_tightpt_str, met_str)
sel_el_phpt_nominal = '%s * ( %s && %s && %s && %s && %s && %s && ( %s ) && ADDITION)' % (
weight_str, sel_base_el, el_tight, el_eta, ph_tightpt_str, met_str,
Zveto_str, el_ip_str)
sel_base_mu = sel_mu_phpt_nominal
sel_base_el = sel_el_phpt_nominal
for ieta in eta_cuts:
#full_suffix = '%s_%s_%s' %(samp.name, suffix, ieta)
name = "_".join(
['MadGraphResonance', 'Mass',
"%d" % mass, 'Width', wid])
full_suffix = '%s_%s_%s' % (name, suffix, ieta)
#eta_str_sr = 'ph_Is%s[%s]' %( ieta, ph_idx_sr )
if 'mu' in suffix:
sel_base = sel_base_mu
extra_label = "Muon channel"
else:
sel_base = sel_base_el
extra_label = "Electron channel"
#full_sel_sr = ' && '.join( [sel_base, ph_selection_sr, eta_str_sr, addtl_cuts_sr] )
full_sel_sr = sel_base.replace(
"ADDITION", " (%s > %d && %s < %d )" %
(plot_var, binning[iwidth][mass][1], plot_var,
binning[iwidth][mass][2]))
print full_sel_sr
#hist_sr = clone_sample_and_draw( sampMan, sampname, plot_var, full_sel_sr, binning )
sampMan.create_hist(samp, plot_var, full_sel_sr,
binning[iwidth][mass])
hist_sr = samp.hist
integral = hist_sr.Integral()
#hist_sr.Scale( 1.0/integral )
#datahist = ROOT.RooDataHist( 'srhist_%s' %full_suffix, 'srhist', ROOT.RooArgList(xvar), hist_sr)
#histpdf = ROOT.RooHistPdf( 'srhistpdf_%s' %(full_suffix), 'srhistpdf' , ROOT.RooArgSet( xvar), datahist, 3 )
assert "mu" in suffix or "el" in suffix, "suffix %s must contain the mu or el channel." % suffix
if 'mu' in suffix:
workspace = ROOT.RooWorkspace(
'workspace_signal_Mass_%d_Width_%s_mu' % (mass, wid))
else:
workspace = ROOT.RooWorkspace(
'workspace_signal_Mass_%d_Width_%s_el' % (mass, wid))
fit_max = mass * 1.20
fit_min = max(mass * 0.50, 200.0)
## set the fit range
xvar.setMin(fit_min)
xvar.setMax(fit_max)
#fitManager = FitManager( 'bwxcb', 0, samp.name, hist_sr, plot_var, ieta, xvar, full_suffix, True,
#sample_params={'mass' : mass, 'width' : width}, )
fitManager = FitManager(
'dscb',
hist=hist_sr,
xvardata=xvar,
label=full_suffix,
sample_params={
'mass': mass,
'width': width
},
)
#fitManager.make_func_pdf()
fitManager.setup_fit()
#fitManager.fit_histogram(workspace )
fitManager.run_fit_minuit(fitrange=(fit_min, fit_max))
fitManager.get_results(workspace)
#fitManager.save_fit( sampMan, workspace, stats_pos='left' , extra_label = extra_label , plotParam =True)
canv = fitManager.draw(subplot="pull",
paramlayout=(0.15, 0.5, 0.82),
useOldsetup=True)
canv.Print("%s/%s.pdf" % (plots_dir, full_suffix))
canv.Print("%s/%s.C" % (plots_dir, full_suffix))
print "************"
print " RooFitResult Status: %d" % fitManager.roofitresult.status()
print "************"
workspaces_to_save.update({workspace.GetName(): workspace})
def makevariableplots_simultaneous(samp,
ptlist,
fitrange,
basesel="1",
tag="",
ic=None,
dobkgd=True,
donorm=True,
maxtimes=5,
doData=False,
var="ph_pt[0]",
useparms=True):
""" extended pdf fit """
if ic is None: ic = dict(sig="dcbp", bkgd="gaus", ext="simul")
if "icparm" not in ic: ic["icparm"] = dict()
parmufloats = defaultdict(list)
parmvals, parmerrs = defaultdict(list), defaultdict(list)
parmnames = FitManager.ParamDCB
parmnames2 = ["gauszg_mean", "gauszg_sig"]
## Get Z+jet/ Zg shape from parameter files
parm2fn = "data/efake/%i/parms_zgamma.txt" % options.year
icondgzg = get_ic(parm2fn, parmnames2, ptlist)
print("icondgzg")
pprint(icondgzg)
parmfn = "data/efake/%i/dcbparms_all%i.txt" % ((options.year, ) * 2)
iconddcb = get_ic(parmfn, parmnames, ptlist)
print("iconddcb")
pprint(iconddcb)
## set up fit manager
fm = FitManager("dcbexpo", xvardata=(50, 250, "GeV"))
for ptrange in zip(ptlist[:-1], ptlist[1:]):
ptrange = tuple(ptrange)
ptname = tuple([str(x).replace('.', 'p') for x in ptrange])
#xp.append(ptrange)
if useparms:
## do some operations here
icdcbtemp = iconddcb[ptrange[0]]
for p in ["dcb_sigma", "dcb_mass"]:
re_parm = getparm(FitManager.setuparray["dcbp"], p)
replaceparm(icdcbtemp, re_parm)
ic['icparm']['dcbp'] = [("x", 50, 250)] + icdcbtemp
ic['icparm']['gauszg'] = [("x", 50, 250)] + icondgzg[ptrange[0]]
### NOTE fitting step
# get default if entry doesnt exist
ptfitrange = fitrange.get(ptrange[0], fitrange[-1])
values, stackcount = fitting_simultaneous(
samp,
fm,
ptrange,
ptfitrange,
basesel=basesel,
var=var,
tag=tag,
ic=ic,
#xbins=(20,50,250),
xbins=[
50,
] + list(range(75, 85, 5)) + list(range(85, 95, 1)) +
list(range(95, 110, 5)) + list(range(110, 250, 10)) + [
250,
],
#xbins = range(50, 251, 1),
dobkgd=dobkgd,
maxtimes=maxtimes,
doData=doData,
donorm=donorm)
if donorm:
make_normalization_comparison(fm, values, stackcount, ptname, tag)
for name, val in list(values.items()):
if val is None: #for case of non-initiated variable
parmvals[name].append(-1)
parmerrs[name].append(-1)
elif val.s > 0: #check if the value is fitted
parmvals[name].append(val.n) #value
parmerrs[name].append(val.s) #uncertainty
parmufloats[name].append(val)
# dealing with width of overflow/last bin
#xpoints, xerrs = xp.output()
print("values:")
pprint(list(parmufloats.items()))
f = lambda p: {x: y for x, y in p.items()}
## output values in json file
data = {'ptlist': ptlist, 'parm': f(parmvals), 'error': f(parmerrs)}
with open('data/efake/%i/parms_%s.txt' % (options.year, tag),
'w') as outfile:
json.dump(data, outfile)
def make_signal_fits( sampMan, sel_base, eta_cuts, plot_var, xvar, binning, workspace, suffix ) :
sampMan.clear_hists()
for samp in sampMan.get_samples(isSignal=True ) :
print 'Sample = ', samp.name
res = re.match('(MadGraph|Pythia)ResonanceMass(\d+)_.*', samp.name )
if res is None :
print 'Could not interpret path ', samp.name
else :
mass = float(res.group(2))
if samp.name.count( 'width0p01' ) :
width = 0.0001
else :
res2 = re.match('(MadGraph|Pythia)ResonanceMass(\d+)_width(\d)', samp.name )
width = float(res2.group(3))/100.
if mass != 450 :
continue
ph_selection_sr = '%s==1' %defs.get_phid_selection('all')
ph_idx_sr = defs.get_phid_idx( 'all' )
print binning[mass]
addtl_cuts_sr = 'ph_pt[%s] > 50 && %s > %f && %s < %f ' %(ph_idx_sr, plot_var, binning[mass][1], plot_var, binning[mass][2] )
xvar.setBins(10000,'cache')
xvar.setMin('cache',-100)
xvar.setMax('cache',1500)
for ieta in eta_cuts :
full_suffix = '%s_%s_%s' %(samp.name, suffix, ieta)
eta_str_sr = 'ph_Is%s[%s]' %( ieta, ph_idx_sr )
full_sel_sr = ' && '.join( [sel_base, ph_selection_sr, eta_str_sr, addtl_cuts_sr] )
#hist_sr = clone_sample_and_draw( sampMan, sampname, plot_var, full_sel_sr, binning )
sampMan.create_hist( samp, plot_var, full_sel_sr, binning[mass] )
hist_sr = samp.hist
integral = hist_sr.Integral()
#hist_sr.Scale( 1.0/integral )
#datahist = ROOT.RooDataHist( 'srhist_%s' %full_suffix, 'srhist', ROOT.RooArgList(xvar), hist_sr)
#histpdf = ROOT.RooHistPdf( 'srhistpdf_%s' %(full_suffix), 'srhistpdf' , ROOT.RooArgSet( xvar), datahist, 3 )
fit_max = mass*1.1
fit_min = mass/1.8
if mass >= 1000 :
fit_max = mass*1.05
fit_min = mass*0.7
xvar.setMin( fit_min )
xvar.setMax( fit_max )
fitManager = FitManager( 'bwxcb', 0, hist_sr, plot_var, ieta, xvar, full_suffix, True,
sample_params={'mass' : mass, 'width' : width}, )
fit_distribution( fitManager)
#save_fit( fitManager, sampMan, workspace, stats_pos='left' )
iter_managers = []
iter_managers.append( fitManager )
saved_result = False
for i in range( 0, 4 ) :
#for i in [0.5, 0.3, 0.2, 0.1, 0.05] :
print 'GOTHERE1'
iter_managers.append(FitManager( 'bwxcb', 0, hist_sr, plot_var, ieta, xvar, full_suffix, True,
sample_params={'mass' : mass, 'width' : width}, ))
print 'GOTHERE2'
cv_sigma = iter_managers[-2].cb_sigma.getValV()
lo_sigma = iter_managers[-2].cb_sigma.getErrorLo()
hi_sigma = iter_managers[-2].cb_sigma.getErrorHi()
cv_power = iter_managers[-2].cb_power.getValV()
lo_power = iter_managers[-2].cb_power.getErrorLo()
hi_power = iter_managers[-2].cb_power.getErrorHi()
cv_mass = iter_managers[-2].cb_mass.getValV()
lo_mass = iter_managers[-2].cb_mass.getErrorLo()
hi_mass = iter_managers[-2].cb_mass.getErrorHi()
err_sigma = hi_sigma/cv_sigma
err_power = hi_power/cv_power
err_mass = hi_mass/cv_mass
#new_lim_sigma = ( cv_sigma, cv_sigma*(1-i), cv_sigma*(1+i))
#new_lim_power = ( cv_power, cv_power*(1-i), cv_power*(1+i) )
#new_lim_mass = ( cv_mass, cv_mass*(1-i), cv_mass*(1+i) )
new_lim_sigma = ( cv_sigma, cv_sigma+lo_sigma, cv_sigma+hi_sigma)
new_lim_power = ( cv_power, cv_power+lo_power, cv_power+hi_power )
new_lim_mass = ( cv_mass, cv_mass+lo_mass, cv_mass+hi_mass )
#if cv_sigma < 0 :
# new_lim_sigma = ( new_lim_sigma[0], new_lim_sigma[2], new_lim_sigma[1])
#if cv_power < 0 :
# new_lim_power = ( new_lim_power[0], new_lim_power[2], new_lim_power[1])
#if cv_mass < 0 :
# new_lim_mass = ( new_lim_mass[0], new_lim_mass[2], new_lim_mass[1] )
print 'NEW DEFAULTS'
print new_lim_sigma
print new_lim_power
print new_lim_mass
iter_managers[-1].set_vals('cb_sigma', mass, new_lim_sigma )
iter_managers[-1].set_vals('cb_power', mass, new_lim_power )
iter_managers[-1].set_vals('cb_mass', mass, new_lim_mass )
print 'GOTHERE3'
fit_distribution( iter_managers[-1] )
print 'GOTHERE4'
cv_sigma_new = iter_managers[-1].cb_sigma.getValV()
lo_sigma_new = iter_managers[-1].cb_sigma.getErrorLo()
hi_sigma_new = iter_managers[-1].cb_sigma.getErrorHi()
cv_power_new = iter_managers[-1].cb_power.getValV()
lo_power_new = iter_managers[-1].cb_power.getErrorLo()
hi_power_new = iter_managers[-1].cb_power.getErrorHi()
cv_mass_new = iter_managers[-1].cb_mass.getValV()
lo_mass_new = iter_managers[-1].cb_mass.getErrorLo()
hi_mass_new = iter_managers[-1].cb_mass.getErrorHi()
err_sigma_new = hi_sigma_new/cv_sigma_new
err_power_new = hi_power_new/cv_power_new
err_mass_new = hi_mass_new/cv_mass_new
print 'Sigma : Previous error = %f, new error = %f' %( err_sigma, err_sigma_new )
print 'Power : Previous error = %f, new error = %f' %( err_power, err_power_new )
print 'Mass : Previous error = %f, new error = %f' %( err_mass, err_mass_new )
print 'GOTHERE5'
# if we get worse results with the new fit, then use the previous one
if math.fabs(err_sigma_new) > math.fabs(err_sigma) or math.fabs(err_power_new) > math.fabs(err_power) or math.fabs(err_mass_new) > math.fabs(err_mass) :
print 'GOTHERE6'
save_fit( iter_managers[-2], sampMan, workspace, stats_pos='left' )
print 'GOTHERE7'
saved_result = True
break
# if we haven't saved the fit yet, then the best
# version is the latest
if not saved_result :
print 'GOTHERE8'
save_fit( iter_managers[-1], sampMan, workspace, stats_pos='left' )
print 'GOTHERE9'
data_n=df.at[0, 'total_test_data'],
)
valid_ds_cls = DataManager(
tfrecord_path=path_cls.get_validation_ds_path(),
img_root=IMAGE_ROOT_PATH,
batch_size=VALIDATION_BATCH_SIZE,
net_cls=net_cls,
data_n=df.at[0, 'total_valid_data'],
)
# ========== ネットワーク呼び出し ==========
fit_cls = FitManager(
net_cls=net_cls,
path_cls=path_cls,
shuf_train_ds_cls=shuf_train_ds_cls,
train_ds_cls=train_ds_cls,
test_ds_cls=test_ds_cls,
validation_ds_cls=valid_ds_cls,
check_img_path=check_img_path,
)
# ========== 学習＆評価 ==========
fit_cls.fit(
end_epoch=END_EPOCH,
device_list=DEVICE_LIST,
ckpt_step=CHECK_STEP,
)
# ========== plot ==========
PlotFunc.accuracy_plot(
path_cls=path_cls,
def make_wjets_fit( sampMan, sample, sel_base, eta_cut, plot_var, shape_var, num_var, binning, xvar, suffix='', closure=False, workspace=None) :
#---------------------------------------
# Get the base selection for each region
#---------------------------------------
ph_selection_sr = '%s==1' %defs.get_phid_selection('medium')
ph_selection_den = '%s==1'%defs.get_phid_selection( num_var, shape_var )
ph_selection_num = '%s==1' %defs.get_phid_selection( num_var )
ph_selection_shape = '%s==1' %defs.get_phid_selection( shape_var )
#ph_selection_ABCD = '%s==1' %defs.get_phid_selection('all')
ph_idx_sr = defs.get_phid_idx( 'medium' )
ph_idx_den = defs.get_phid_idx( num_var, shape_var )
ph_idx_num = defs.get_phid_idx( num_var )
ph_idx_shape = defs.get_phid_idx( shape_var )
#ph_idx_ABCD = defs.get_phid_idx('all')
xmin = xvar.getMin()
xmax = xvar.getMax()
#---------------------------------------
# Add eta cuts, (IsEB, IsEE)
#---------------------------------------
eta_str_shape = 'ph_Is%s[%s]' %( eta_cut, ph_idx_shape )
eta_str_den = 'ph_Is%s[%s]' %( eta_cut, ph_idx_den)
eta_str_num = 'ph_Is%s[%s]' %( eta_cut, ph_idx_num )
eta_str_sr = 'ph_Is%s[%s]' %( eta_cut, ph_idx_sr )
#eta_str_ABCD = 'ph_Is%s[%s]' %( eta_cut, ph_idx_ABCD )
#---------------------------------------
# Add additional cuts, mainly restricting
# the fitted variable to the plotting limits
#---------------------------------------
addtl_cuts_sr = 'ph_pt[%s] > 50 && %s > %d && %s < %d ' %( ph_idx_sr, plot_var, xmin, plot_var, xmax )
addtl_cuts_den = 'ph_pt[%s] > 50 && %s > %d && %s < %d ' %( ph_idx_den, plot_var, xmin, plot_var, xmax )
addtl_cuts_num = 'ph_pt[%s] > 50 && %s > %d && %s < %d ' %( ph_idx_num, plot_var, xmin, plot_var, xmax )
addtl_cuts_shape = 'ph_pt[%s] > 50 && %s > %d && %s < %d ' %( ph_idx_shape, plot_var, xmin, plot_var, xmax )
#addtl_cuts_ABCD = 'ph_pt[%s] > 50 && %s > %d && %s < %d ' %( ph_idx_ABCD, plot_var, xmin, plot_var, xmax )
#---------------------------------------
# Get the cuts that define the photon
# sideband regions
#---------------------------------------
cut_str_base = ' {var}[{idx}] > {val_low} && {var}[{idx}] < {val_high}'
cut_var_shape = defs.get_phid_cut_var( shape_var )
cut_var_num = defs.get_phid_cut_var( num_var )
cut_vals_shape = get_cut_defaults( shape_var, eta_cut )
cut_vals_num = get_cut_defaults( num_var, eta_cut )
cut_str_shape = cut_str_base.format(var=cut_var_shape, idx=ph_idx_shape, val_low=cut_vals_shape[0], val_high=cut_vals_shape[1] )
cut_str_num = cut_str_base.format(var=cut_var_num, idx=ph_idx_num, val_low=cut_vals_num[0], val_high=cut_vals_num[1] )
cut_str_den_1 = cut_str_base.format(var=cut_var_shape, idx=ph_idx_den, val_low=cut_vals_shape[0], val_high=cut_vals_shape[1] )
cut_str_den_2 = cut_str_base.format(var=cut_var_num, idx=ph_idx_den, val_low=cut_vals_num[0], val_high=cut_vals_num[1] )
cut_str_den = cut_str_den_1 + ' && ' + cut_str_den_2
#---------------------------------------
# put the cuts together
#---------------------------------------
full_sel_shape = ' && '.join( [sel_base, ph_selection_shape, eta_str_shape, addtl_cuts_shape, cut_str_shape ] )
full_sel_num = ' && '.join( [sel_base, ph_selection_num, eta_str_num, addtl_cuts_num, cut_str_num] )
full_sel_den = ' && '.join( [sel_base, ph_selection_den, eta_str_den, addtl_cuts_den, cut_str_den] )
full_sel_sr = ' && '.join( [sel_base, ph_selection_sr, eta_str_sr, addtl_cuts_sr] )
#full_sel_ABCD = ' && '.join( [sel_base, ph_selection_ABCD, eta_str_ABCD, addtl_cuts_ABCD] )
label_shape = 'shape_%s_%s' %(sample,suffix)
label_num = 'num_%s_%s' %(sample,suffix)
label_den = 'den_%s_%s' %(sample,suffix)
label_sr = 'sr_%s_%s' %(sample,suffix)
if workspace is None :
ws = ROOT.RooWorkspace( 'ws')
else :
ws = workspace
jet_var = 'leadjet_pt' # 0 to 640 in bins of 5
jet_binning = (128,0.,640.)
subjet_var = 'subljet_pt' # 0 to 380 in bins of 3.8
subjet_binning = (100,0.,380.)
ht_var = 'trueht' # 0 to 7000 in bins of 50
ht_binning = (140,0.,7000.)
ABCD_var = 'ph_sigmaIEIE:ph_chIsoCorr'
ABCD_binning = (120,0.,12.,50,0.,0.05)
#chIsovar = 'ph_chIso'
#sigIEIEvar = 'ph_sigmaIEIE'
chIso_B_var = 'ph_chIsoCorr[ptSorted_ph_mediumNoChIso_idx[0]]'
sigIEIE_B_var = 'ph_sigmaIEIE[ptSorted_ph_mediumNoChIso_idx[0]]'
chIso_C_var = 'ph_chIsoCorr[ptSorted_ph_mediumNoSIEIE_idx[0]]'
sigIEIE_C_var = 'ph_sigmaIEIE[ptSorted_ph_mediumNoSIEIE_idx[0]]'
chIso_D_var = 'ph_chIsoCorr[ptSorted_ph_mediumNoSIEIENoChIso_idx[0]]'
sigIEIE_D_var = 'ph_sigmaIEIE[ptSorted_ph_mediumNoSIEIENoChIso_idx[0]]'
twoD_B_var = 'ph_sigmaIEIE[ptSorted_ph_mediumNoChIso_idx[0]]:ph_chIsoCorr[ptSorted_ph_mediumNoChIso_idx[0]]'
twoD_C_var = 'ph_sigmaIEIE[ptSorted_ph_mediumNoSIEIE_idx[0]]:ph_chIsoCorr[ptSorted_ph_mediumNoSIEIE_idx[0]]'
twoD_D_var = 'ph_sigmaIEIE[ptSorted_ph_mediumNoSIEIENoChIso_idx[0]]:ph_chIsoCorr[ptSorted_ph_mediumNoSIEIENoChIso_idx[0]]'
sigIEIE_binning = (50,0.,0.05)
chIso_binning = (120,0.,12.)
#---------------------------------------
# draw the histograms
#---------------------------------------
hist_shape = clone_sample_and_draw( sampMan, sample, plot_var, full_sel_shape, binning )
hist_num = clone_sample_and_draw( sampMan, sample, plot_var, full_sel_num , binning )
hist_den = clone_sample_and_draw( sampMan, sample, plot_var, full_sel_den , binning )
sampMan.outputs['%s_shapehist_%s' %(sample,suffix)] = hist_shape
sampMan.outputs['%s_numhist_%s' %(sample,suffix)] = hist_num
sampMan.outputs['%s_denhist_%s' %(sample,suffix)] = hist_den
#hist_ABCD = clone_sample_and_draw( sampMan, sample, ABCD_var, full_sel_ABCD, ABCD_binning )
hist_B = clone_sample_and_draw( sampMan, sample, twoD_B_var, full_sel_shape, ABCD_binning )
hist_C = clone_sample_and_draw( sampMan, sample, twoD_C_var, full_sel_num, ABCD_binning )
hist_D = clone_sample_and_draw( sampMan, sample, twoD_D_var, full_sel_den, ABCD_binning )
hist_1d_sigmaIEIE_B = clone_sample_and_draw( sampMan, sample, sigIEIE_B_var, full_sel_shape, sigIEIE_binning )
hist_1d_chIso_B = clone_sample_and_draw( sampMan, sample, chIso_B_var, full_sel_shape, chIso_binning )
hist_1d_sigmaIEIE_C = clone_sample_and_draw( sampMan, sample, sigIEIE_C_var, full_sel_num, sigIEIE_binning )
hist_1d_chIso_C = clone_sample_and_draw( sampMan, sample, chIso_C_var, full_sel_num, chIso_binning )
hist_1d_sigmaIEIE_D = clone_sample_and_draw( sampMan, sample, sigIEIE_D_var, full_sel_den, sigIEIE_binning )
hist_1d_chIso_D = clone_sample_and_draw( sampMan, sample, chIso_D_var, full_sel_den, chIso_binning )
sampMan.outputs['%s_sigmaIEIE_B_%s' %(sample,suffix)] = hist_1d_sigmaIEIE_B
sampMan.outputs['%s_sigmaIEIE_C_%s' %(sample,suffix)] = hist_1d_sigmaIEIE_C
sampMan.outputs['%s_sigmaIEIE_D_%s' %(sample,suffix)] = hist_1d_sigmaIEIE_D
sampMan.outputs['%s_chIso_B_%s' %(sample,suffix)] = hist_1d_chIso_B
sampMan.outputs['%s_chIso_C_%s' %(sample,suffix)] = hist_1d_chIso_C
sampMan.outputs['%s_chIso_D_%s' %(sample,suffix)] = hist_1d_chIso_D
#---------------------------------------
# make fit managers
#---------------------------------------
fitMan_shape = FitManager( 'dijet', 3, '%s_shape' %sample, hist_shape, plot_var, eta_cut, xvar, label_shape, useRooFit=False )
fitMan_num = FitManager( 'dijet', 3, '%s_num' %sample, hist_num, plot_var, eta_cut, xvar, label_num, useRooFit=False )
fitMan_den = FitManager( 'dijet', 3, '%s_den' %sample, hist_den, plot_var, eta_cut, xvar, label_den, useRooFit=False )
#---------------------------------------
# Do the fits
#---------------------------------------
result_shape= fitMan_shape.fit_histogram( ws )
result_num = fitMan_num.fit_histogram( ws )
result_den = fitMan_den.fit_histogram( ws )
print 'DEBUG: results of fit manager (shape, num, den): ',result_shape,result_num,result_den
#---------------------------------------
# save the results
#---------------------------------------
fitMan_den.save_fit( sampMan, ws, logy=True )
fitMan_shape.save_fit( sampMan, ws, logy=True )
fitMan_num.save_fit( sampMan, ws, logy=True )
#---------------------------------------
# calculate the function that describes
# the ratio of num/den
#---------------------------------------
_integral_num = ws.pdf( 'dijet_%s' %label_num ).getNormIntegral(ROOT.RooArgSet( xvar ) )
_integral_den = ws.pdf( 'dijet_%s' %label_den).getNormIntegral(ROOT.RooArgSet( xvar ) )
_integral_shape = ws.pdf( 'dijet_%s' %label_shape).getNormIntegral(ROOT.RooArgSet( xvar ) )
func_integral_num = _integral_num.getValV()
func_integral_den = _integral_den.getValV()
func_integral_shape = _integral_shape.getValV()
hist_integral_num = result_num['integral'].n
hist_integral_den = result_den['integral'].n
hist_integral_shape = result_shape['integral'].n
norm_num = hist_integral_num / func_integral_num
norm_den = hist_integral_den / func_integral_den
norm_shape = hist_integral_shape / func_integral_shape
print 'func integral Num = ', func_integral_num
print 'hist integral Num = ', hist_integral_num
print 'normalization Num = ', norm_num
print 'func integral Den = ', func_integral_den
print 'hist integral Den = ', hist_integral_den
print 'normalization Den = ', norm_den
print 'func integral Shape = ', func_integral_shape
print 'hist integral Shape = ', hist_integral_shape
print 'normalization Shape = ', norm_shape
power_pred_name = 'power_pred_%s' %suffix
logcoef_pred_name = 'logcoef_pred_%s' %suffix
log2coef_pred_name = 'log2coef_pred_%s' %suffix
power_ratio_name = 'power_ratio_%s' %suffix
logcoef_ratio_name = 'logcoef_ratio_%s' %suffix
#name_power_num ='power'
#name_power_den ='power'
#name_power_shape ='power'
#name_logcoef_num ='logcoef1'
#name_logcoef_den ='logcoef1'
#name_logcoef_shape ='logcoef1'
name_power_num ='dijet_order1_%s' %label_num
name_power_den ='dijet_order1_%s' %label_den
name_power_shape ='dijet_order1_%s' %label_shape
name_logcoef_num ='dijet_order2_%s' %label_num
name_logcoef_den ='dijet_order2_%s' %label_den
name_logcoef_shape ='dijet_order2_%s' %label_shape
name_log2coef_num ='dijet_order3_%s' %label_num
name_log2coef_den ='dijet_order3_%s' %label_den
name_log2coef_shape ='dijet_order3_%s' %label_shape
val_power_num = ws.var( name_power_num )
val_power_den = ws.var( name_power_den )
val_power_shape = ws.var( name_power_shape )
val_logcoef_num = ws.var( name_logcoef_num )
val_logcoef_den = ws.var( name_logcoef_den )
val_logcoef_shape = ws.var( name_logcoef_shape )
val_log2coef_num = ws.var( name_log2coef_num )
val_log2coef_den = ws.var( name_log2coef_den )
val_log2coef_shape = ws.var( name_log2coef_shape )
power_pred = ROOT.RooRealVar( power_pred_name , 'power' , (val_power_num.getValV() + val_power_shape.getValV() - val_power_den.getValV()) , -100, 100)
logcoef_pred = ROOT.RooRealVar( logcoef_pred_name , 'logcoef', (val_logcoef_num.getValV() + val_logcoef_shape.getValV() - val_logcoef_den.getValV()), -10, 10 )
log2coef_pred = ROOT.RooRealVar( log2coef_pred_name , 'log2coef', (val_log2coef_num.getValV() + val_log2coef_shape.getValV() - val_log2coef_den.getValV()), -10, 10 )
#power_ratio = ROOT.RooRealVar( power_ratio_name , 'power' , val_power_num - val_power_den , -100, 100)
#logcoef_ratio = ROOT.RooRealVar( logcoef_ratio_name, 'logcoef', val_logcoef_num - val_logcoef_den, -10, 10 )
#func = 'TMath::Power(@0/13000, @1+@2*TMath::Log10(@0/13000))'
#prediction = ROOT.RooGenericPdf('dijet_prediction_%s' %suffix , 'prediction', func, ROOT.RooArgList(xvar,power_pred, logcoef_pred))
func = 'TMath::Power(@0/13000, @1+@2*TMath::Log10(@0/13000)+@3*TMath::Log10(@0/13000)*TMath::Log10(@0/13000))'
prediction = ROOT.RooGenericPdf('dijet_prediction_%s' %suffix , 'prediction', func, ROOT.RooArgList(xvar,power_pred, logcoef_pred, log2coef_pred))
norm_pred = ROOT.RooRealVar( 'dijet_prediction_%s_norm' %(suffix), 'prediction normalization', (hist_integral_shape * hist_integral_num) / hist_integral_den )
getattr( ws , 'import' ) ( norm_pred )
getattr( ws , 'import' ) ( prediction )
#---------------------------------------
# Save the ABCD plots
#---------------------------------------
#can_ABCD = ROOT.TCanvas( str(uuid.uuid4()), '' )
#can_ABCD.SetLogz()
#hist_ABCD.Scale(1./hist_ABCD.Integral())
#hist_ABCD.Draw("colz")
#print "DEBUG: ABCD integral = ",hist_ABCD.Integral()
can_B = ROOT.TCanvas( str(uuid.uuid4()), '' )
can_B.SetLogz()
hist_B.Scale(1./hist_B.Integral())
hist_B.Draw("colz")
print "DEBUG: B integral = ",hist_B.Integral()
can_C = ROOT.TCanvas( str(uuid.uuid4()), '' )
hist_C.Scale(1./hist_C.Integral())
can_C.SetLogz()
hist_C.Draw("colz")
print "DEBUG: C integral = ",hist_C.Integral()
can_D = ROOT.TCanvas( str(uuid.uuid4()), '' )
hist_D.Scale(1./hist_D.Integral())
can_D.SetLogz()
hist_D.Draw("colz")
print "DEBUG: D integral = ",hist_D.Integral()
can_1d_sig_B = ROOT.TCanvas( str(uuid.uuid4()), '' )
can_1d_sig_B.SetLogy()
hist_1d_sigmaIEIE_B.Draw('l')
can_1d_iso_B = ROOT.TCanvas( str(uuid.uuid4()), '' )
can_1d_iso_B.SetLogy()
hist_1d_chIso_B.Draw('l')
can_1d_sig_C = ROOT.TCanvas( str(uuid.uuid4()), '' )
can_1d_sig_C.SetLogy()
hist_1d_sigmaIEIE_C.Draw('l')
can_1d_iso_C = ROOT.TCanvas( str(uuid.uuid4()), '' )
can_1d_iso_C.SetLogy()
hist_1d_chIso_C.Draw('l')
can_1d_sig_D = ROOT.TCanvas( str(uuid.uuid4()), '' )
can_1d_sig_D.SetLogy()
hist_1d_sigmaIEIE_D.Draw('l')
can_1d_iso_D = ROOT.TCanvas( str(uuid.uuid4()), '' )
can_1d_iso_D.SetLogy()
hist_1d_chIso_D.Draw('l')
#plot sigmaIEIE in regions B, C, D
can_1d_sig = ROOT.TCanvas( str(uuid.uuid4()), '' )
can_1d_sig.SetLogy()
hist_norm_1d_sigmaIEIE_B = ROOT.TH1F(str(uuid.uuid4()) , str(uuid.uuid4()), hist_1d_sigmaIEIE_B.GetNbinsX(), hist_1d_sigmaIEIE_B.GetXaxis().GetXmin(), hist_1d_sigmaIEIE_B.GetXaxis().GetXmax())
hist_norm_1d_sigmaIEIE_C = ROOT.TH1F(str(uuid.uuid4()) , str(uuid.uuid4()), hist_1d_sigmaIEIE_C.GetNbinsX(), hist_1d_sigmaIEIE_C.GetXaxis().GetXmin(), hist_1d_sigmaIEIE_C.GetXaxis().GetXmax())
hist_norm_1d_sigmaIEIE_D = ROOT.TH1F(str(uuid.uuid4()) , str(uuid.uuid4()), hist_1d_sigmaIEIE_D.GetNbinsX(), hist_1d_sigmaIEIE_D.GetXaxis().GetXmin(), hist_1d_sigmaIEIE_D.GetXaxis().GetXmax())
hist_1d_sigmaIEIE_B.Copy(hist_norm_1d_sigmaIEIE_B)
hist_1d_sigmaIEIE_C.Copy(hist_norm_1d_sigmaIEIE_C)
hist_1d_sigmaIEIE_D.Copy(hist_norm_1d_sigmaIEIE_D)
ROOT.SetOwnership( hist_norm_1d_sigmaIEIE_B, False )
ROOT.SetOwnership( hist_norm_1d_sigmaIEIE_C, False )
ROOT.SetOwnership( hist_norm_1d_sigmaIEIE_D, False )
hist_norm_1d_sigmaIEIE_B.Scale(1./hist_norm_1d_sigmaIEIE_B.Integral())
hist_norm_1d_sigmaIEIE_C.Scale(1./hist_norm_1d_sigmaIEIE_C.Integral())
hist_norm_1d_sigmaIEIE_D.Scale(1./hist_norm_1d_sigmaIEIE_D.Integral())
hist_norm_1d_sigmaIEIE_B.SetLineColor(2)
hist_norm_1d_sigmaIEIE_C.SetLineColor(1)
hist_norm_1d_sigmaIEIE_D.SetLineColor(4)
hist_norm_1d_sigmaIEIE_B.SetMarkerColor(2)
hist_norm_1d_sigmaIEIE_C.SetMarkerColor(1)
hist_norm_1d_sigmaIEIE_D.SetMarkerColor(4)
hist_norm_1d_sigmaIEIE_B.Draw('l')
hist_norm_1d_sigmaIEIE_C.Draw('same')
hist_norm_1d_sigmaIEIE_D.Draw('same')
#plot chIso in regions B, C, D
can_1d_iso = ROOT.TCanvas( str(uuid.uuid4()), '' )
can_1d_iso.SetLogy()
hist_norm_1d_chIso_B = ROOT.TH1F(str(uuid.uuid4()) , str(uuid.uuid4()), hist_1d_chIso_B.GetNbinsX(), hist_1d_chIso_B.GetXaxis().GetXmin(), hist_1d_chIso_B.GetXaxis().GetXmax())
hist_norm_1d_chIso_C = ROOT.TH1F(str(uuid.uuid4()) , str(uuid.uuid4()), hist_1d_chIso_C.GetNbinsX(), hist_1d_chIso_C.GetXaxis().GetXmin(), hist_1d_chIso_C.GetXaxis().GetXmax())
hist_norm_1d_chIso_D = ROOT.TH1F(str(uuid.uuid4()) , str(uuid.uuid4()), hist_1d_chIso_D.GetNbinsX(), hist_1d_chIso_D.GetXaxis().GetXmin(), hist_1d_chIso_D.GetXaxis().GetXmax())
hist_1d_chIso_B.Copy(hist_norm_1d_chIso_B)
hist_1d_chIso_C.Copy(hist_norm_1d_chIso_C)
hist_1d_chIso_D.Copy(hist_norm_1d_chIso_D)
ROOT.SetOwnership( hist_norm_1d_chIso_B, False )
ROOT.SetOwnership( hist_norm_1d_chIso_C, False )
ROOT.SetOwnership( hist_norm_1d_chIso_D, False )
hist_norm_1d_chIso_B.Scale(1./hist_norm_1d_chIso_B.Integral())
hist_norm_1d_chIso_C.Scale(1./hist_norm_1d_chIso_C.Integral())
hist_norm_1d_chIso_D.Scale(1./hist_norm_1d_chIso_D.Integral())
hist_norm_1d_chIso_B.SetLineColor(2)
hist_norm_1d_chIso_C.SetLineColor(1)
hist_norm_1d_chIso_D.SetLineColor(4)
hist_norm_1d_chIso_B.SetMarkerColor(2)
hist_norm_1d_chIso_C.SetMarkerColor(1)
hist_norm_1d_chIso_D.SetMarkerColor(4)
hist_norm_1d_chIso_C.Draw('l')
hist_norm_1d_chIso_B.Draw('same')
hist_norm_1d_chIso_D.Draw('same')
#---------------------------------------
# Save the ratio histogram
#---------------------------------------
# canvas for ratio histogram
can_ratio = ROOT.TCanvas( 'canv_ratio_%s' %(plot_var), 'canv_ratio_%s' %(plot_var) )
#ROOT.SetOwnership( can_ratio, False )
# ratio histogram = numerator histogram/denominator histogram
ratiohist = ROOT.TH1F( 'ratio_%s' %(plot_var), 'ratio_%s' %(plot_var), hist_num.GetNbinsX(), hist_num.GetXaxis().GetXmin(), hist_num.GetXaxis().GetXmax())
hist_num.Copy( ratiohist)
ROOT.SetOwnership( ratiohist, False )
print "DEBUG: first bin of num before div = ",ratiohist.GetBinContent(1)
print "DEBUG: first bin of den before div = ",hist_den.GetBinContent(1)
ratiohist.Divide( hist_den )
print "DEBUG: first bin of ratio after div = ",ratiohist.GetBinContent(1)
ratiohist.SetMarkerStyle(20)
ratiohist.SetMarkerSize(1)
ratiohist.GetXaxis().SetTitle( 'Transverse Mass [GeV]' )
ratiohist.GetYaxis().SetTitle( 'ratio of passing to failing %s' %shape_var )
ratiohist.Draw()
ratiohist.SetStats(0)
ratiohist.SetMinimum( 0)
ratiohist.SetMaximum( 5)
#ratio_func = ROOT.TF1( 'ratio_func_%s'%(str(uuid.uuid4())), '( [2]*TMath::Power(x/13000, [0] + [1]*TMath::Log10(x/13000) ) ) ', xmin, xmax )
ratio_func = ROOT.TF1( 'ratio_func', '[0]', xmin, xmax )
print 'DEBUG: ratio function = ',ratio_func
ROOT.SetOwnership( ratio_func, False )
print 'DEBUG: ratio function value at 200 GeV = ',ratio_func.Eval(200.)
print 'DEBUG: ratio of integrals = ',hist_num.Integral()/hist_den.Integral()
ratio_func.SetParameter(0, hist_num.Integral()/hist_den.Integral())
#ratio_func.SetParameter(0, result_num[name_power_num].n - result_den[name_power_den].n)
#ratio_func.SetParameter(1, result_num[name_logcoef_num].n - result_den[name_logcoef_den].n )
#ratio_func.SetParameter(2, norm_num / norm_den )
ratiohist.Fit("ratio_func","RL","",xmin,xmax)
print 'xmin and xmax are ',xmin,xmax
f = ratiohist.GetFunction("ratio_func")
fitvalue = ratio_func.GetParameter(0)
print 'DEBUG: fitted straight line at ',fitvalue
print 'DEBUG: chisquare of fit = ',f.GetChisquare()
print 'DEBUG: ndof of fit = ',f.GetNDF()
print 'DEBUG: testing f->GetParameter(0) = ',f.GetParameter(0)
print 'DEBUG: testing f->GetParError(0) = ',f.GetParError(0)
ratio_func.Draw('same')
print 'DEBUG: prediction in A from B*C/D histograms = ',hist_shape.Integral(hist_shape.FindBin( xmin ), hist_shape.FindBin( xmax ))*hist_num.Integral(hist_num.FindBin( xmin ), hist_num.FindBin( xmax ))/hist_den.Integral(hist_den.FindBin( xmin ), hist_den.FindBin( xmax ))
#ratio_power_v = val_power_num.getValV() - val_power_den.getValV()
#ratio_power_e = math.sqrt(val_power_num.getErrorHi()*val_power_num.getErrorHi() + val_power_den.getErrorHi()*val_power_den.getErrorHi() )
#ratio_logcoef_v = val_logcoef_num.getValV() - val_logcoef_den.getValV()
#ratio_logcoef_e = math.sqrt(val_logcoef_num.getErrorHi()*val_logcoef_num.getErrorHi() + val_logcoef_den.getErrorHi()*val_logcoef_den.getErrorHi() )
#power_tex = ROOT.TLatex(0, 0, 'power = %.01f #pm %.02f' %( ratio_power_v, ratio_power_e ))
#ROOT.SetOwnership( power_tex, False )
#logcoef_tex = ROOT.TLatex(0, 0, 'logcoef = %.01f #pm %.02f' %( ratio_logcoef_v, ratio_logcoef_e ))
#ROOT.SetOwnership( logcoef_tex, False )
#power_tex.SetNDC()
#logcoef_tex.SetNDC()
#power_tex .SetX( 0.6 )
#power_tex .SetY( 0.84 )
#logcoef_tex.SetX( 0.6 )
#logcoef_tex.SetY( 0.78 )
#power_tex.Draw('same')
#logcoef_tex.Draw('same')
#sampMan.outputs['wjets_ABCD_%s' %suffix] = can_ABCD
sampMan.outputs['%s_B_%s' %(sample,suffix)] = can_B
sampMan.outputs['%s_C_%s' %(sample,suffix)] = can_C
sampMan.outputs['%s_D_%s' %(sample,suffix)] = can_D
sampMan.outputs['%s_sigmaIEIE_B_%s' %(sample,suffix)] = can_1d_sig_B
sampMan.outputs['%s_chIso_B_%s' %(sample,suffix)] = can_1d_iso_B
sampMan.outputs['%s_sigmaIEIE_C_%s' %(sample,suffix)] = can_1d_sig_C
sampMan.outputs['%s_chIso_C_%s' %(sample,suffix)] = can_1d_iso_C
sampMan.outputs['%s_sigmaIEIE_D_%s' %(sample,suffix)] = can_1d_sig_D
sampMan.outputs['%s_chIso_D_%s' %(sample,suffix)] = can_1d_iso_D
sampMan.outputs['%s_sigmaIEIE_BCD_%s' %(sample,suffix)] = can_1d_sig
sampMan.outputs['%s_chIso_BCD_%s' %(sample,suffix)] = can_1d_iso
if closure :
# sampMan.outputs['wjetsclosure_ratio_%s' %suffix] = can_ratio
sampMan.outputs['%sclosure_ratio_hist_%s' %(sample,suffix)] = ratiohist
sampMan.outputs['%sclosure_ratio_func_%s' %(sample,suffix)] = ratio_func
#sampMan.outputs['wjetsclosure_ratio_power_%s' %suffix] = power_tex
#sampMan.outputs['wjetsclosure_ratio_logcoef_%s' %suffix] = logcoef_tex
else :
# sampMan.outputs['wjets_ratio_%s' %suffix] = can_ratio
sampMan.outputs['%s_ratio_hist_%s' %(sample,suffix)] = ratiohist
sampMan.outputs['%s_ratio_func_%s' %(sample,suffix)] = ratio_func
#sampMan.outputs['wjets_ratio_power_%s' %suffix] = power_tex
#sampMan.outputs['wjets_ratio_logcoef_%s' %suffix] = logcoef_tex
if closure :
# canvas for signal region predictions
can_sr = ROOT.TCanvas( 'canv_sig_%s_%s' %(sample,plot_var), 'canv_sig_%s_%s' %(sample,plot_var) )
ROOT.SetOwnership( can_sr, False )
hist_sr = clone_sample_and_draw( sampMan, sample, plot_var, full_sel_sr , binning )
ROOT.SetOwnership( hist_sr, False )
hist_sr.GetXaxis().SetTitle( 'Transverse Mass [GeV]' )
#hist_sr.GetYaxis().SetTitle( 'ratio of passing to failing %s' %shape_var ) ##really?
hist_sr.GetYaxis().SetTitle( 'Events in signal region' )
# fit a dijet2 function to region A
fitMan_sr = FitManager( 'dijet', 3, sample, hist_sr, plot_var, eta_cut, xvar, suffix, useRooFit=False )
result_sr = fitMan_sr.fit_histogram( ws )
# from histograms in MC regions B, C, and D
# predicted number of sr events
pred_val = (hist_integral_shape * hist_integral_num) / hist_integral_den
# from histogram in MC region A
# predicted number of sr events
tot_sr = hist_sr.Integral( hist_sr.FindBin( xmin ), hist_sr.FindBin( xmax ) )
print 'DEBUG: SR prediction from BCD MC = ',pred_val
print 'DEBUG: SR MC events from A = ',tot_sr
print 'SR integral = ', tot_sr
# from functions fitted to regions BCD
# predicted number of sr events
sr_func = ROOT.TF1( 'sr_func', '( [2]*TMath::Power(x/13000, [0] + [1]*TMath::Log10(x/13000) ) ) ', xmin, xmax )
#sr_func = ROOT.TF1( 'sr_func', '[3]*TMath::Power(x/13000, [0]+[1]*TMath::Log10(x/13000)+[2]*TMath::Log10(x/13000)*TMath::Log10(x/13000))', xmin, xmax )
ROOT.SetOwnership( sr_func, False )
sr_func.SetParameter(0, result_num[name_power_num].n + result_shape[name_power_shape].n - result_den[name_power_den].n )
sr_func.SetParameter(1, result_num[name_logcoef_num].n + result_shape[name_logcoef_shape].n - result_den[name_logcoef_den].n )
#sr_func.SetParameter(2, result_num[name_log2coef_num].n + result_shape[name_log2coef_shape].n - result_den[name_log2coef_den].n )
sr_func.SetParameter(2, 1)
sr_int = sr_func.Integral( xmin, xmax )
print 'normalizations (shape, num, den) = ',norm_shape,norm_num,norm_den
#print 'Normalization = ',(norm_shape*norm_num) / norm_den
print 'func Integral SR Before = ', sr_func.Integral( xmin, xmax )
sr_func.SetParameter(2, pred_val/sr_int )
print 'func Integral SR After = ', sr_func.Integral( xmin, xmax )
print 'DEBUG: closure SR fit function value at 200 GeV = ',sr_func.Eval(200.)
hist_sr.Draw()
sr_func.Draw('same')
sampMan.outputs['%sclosure_sr_pred_%s' %(sample,suffix)] = can_sr
sampMan.outputs['%sclosure_srhist_%s' %(sample,suffix)] = hist_sr
tot_ratio = result_num['integral']/result_den['integral']
print 'DEBUG: result_num[integral]/result_den[integral] = ',tot_ratio
hist_jetpt = clone_sample_and_draw( sampMan, sample, jet_var, sel_base, jet_binning )
hist_subjetpt = clone_sample_and_draw( sampMan, sample, subjet_var, sel_base, subjet_binning )
hist_trueht = clone_sample_and_draw( sampMan, sample, ht_var, sel_base, ht_binning )
sampMan.outputs['%sclosure_jetpt_%s' %(sample,suffix)] = hist_jetpt
sampMan.outputs['%sclosure_subjetpt_%s' %(sample,suffix)] = hist_subjetpt
sampMan.outputs['%sclosure_trueht_%s' %(sample,suffix)] = hist_trueht
def get_mc_fit(sampMan,
sampnames,
sel_base,
eta_cuts,
xvar,
plot_var,
binning,
workspace,
extra_label="Muon Channel",
suffix='',
plots_dir="plots"):
if not os.path.isdir(plots_dir):
os.makedirs(plots_dir)
print "\n *****************\n calling get_mc_fit for, ", tPurple % (
sampnames), "\n *********************\n"
xmin = xvar.getMin()
xmax = xvar.getMax()
results = {}
full_sel_sr = sel_base.replace(
"ADDITION",
" (%s > %d && %s < %d )" % (plot_var, xmin, plot_var, xmax))
print full_sel_sr
outname = "Backgrounds"
if isinstance(sampnames, str):
outname = sampnames
sampnames = [sampnames]
#hist_sr = clone_sample_and_draw( sampMan, sampname, plot_var, full_sel_sr , binning )
print sampnames
# list of samples
hist_sr = clone_sample_and_draw(sampMan, sampnames[0], plot_var,
full_sel_sr, binning)
for isamp in sampnames[1:]:
hist_sr.Add(
clone_sample_and_draw(sampMan, isamp, plot_var, full_sel_sr,
binning))
print " **** sampname %s number of total events %f **********" % (
outname, hist_sr.Integral(0, 100000))
print " **** sampname %s number of events %f **********" % (
outname, hist_sr.Integral())
ieta = "EB"
fitfunc = "expow"
label = '%s_%s_%s_%s' % (outname, suffix, ieta, fitfunc)
#fitManager = FitManager( 'dijet', 3, sampnames, hist_sr, plot_var, ieta, xvar, label, False)
if fitfunc == "dijet":
fitManager = FitManager('dijet',
hist=hist_sr,
xvardata=xvar,
label=label,
norders=2)
#fitManager = FitManager( 'power', 2, sampnames, hist_sr, plot_var, ieta, xvar, label, options.useRooFit)
if fitfunc == "power":
fitManager = FitManager('power', hist_sr, xvar, label, norders=1)
if fitfunc == "expow":
fitManager = FitManager('expow', hist_sr, xvar, label, norders=1)
if fitfunc == "vvdijet":
fitManager = FitManager('vvdijet', hist_sr, xvar, label, norders=1)
if fitfunc == "atlas":
fitManager = FitManager('atlas', hist_sr, xvar, label, norders=1)
canv = fitManager.draw(paramlayout=(0.7, 0.5, 0.82),
useOldsetup=True,
logy=1,
yrange=(5e-3, 2e4))
#fitManager.setup_fit()
fitManager.setup_rootfit((xmin, xmax))
fitManager.func.Draw("same")
canv.Print("%s/%sbefore.pdf" % (plots_dir, label))
canv.Print("%s/%sbefore.png" % (plots_dir, label))
canv.Print("%s/%sbefore.C" % (plots_dir, label))
fitManager.run_rootfit()
#fit_distribution( fitManager, sampMan, workspace, logy=True )
#fitManager.make_func_pdf()
#fitManager.fit_histogram( workspace )
""" cast it from TF1 to RooGenericPdf """
fitManager.calculate_func_pdf()
fitManager.get_results(workspace)
#results[ieta] = save_distribution( fitManager, sampMan, workspace, logy=True )
#fitManager.save_fit( sampMan, workspace, logy = True, stats_pos='right', extra_label = extra_label)
canv = fitManager.draw(subplot="pull",
paramlayout=(0.7, 0.5, 0.82),
useOldsetup=True,
logy=1,
yrange=(5e-3, 2e4))
canv.Print("%s/%s.pdf" % (plots_dir, label))
canv.Print("%s/%s.png" % (plots_dir, label))
canv.Print("%s/%s.C" % (plots_dir, label))
return results