Exemple #1
0
def latexfitresults_method2(filename,
                            resultname='RooExpandedFitResult_afterFit',
                            region='3jL',
                            sample='',
                            fitregions='WR,TR,S3,S4,SR3jT,SR4jT',
                            dataname='obsData'):

    #  namemap = {}
    #  namemap = getnamemap()

    ############################################

    w = Util.GetWorkspaceFromFile(filename, 'w')
    if w == None:
        print "ERROR : Cannot open workspace : "
        sys.exit(1)

    result = w.obj(resultname)
    if result == None:
        print "ERROR : Cannot open fit result : ", resultname
        sys.exit(1)

    resultlistOrig = result.floatParsFinal()

    snapshot = 'snapshot_paramsVals_' + resultname
    w.loadSnapshot(snapshot)

    data_set = w.data(dataname)
    if data_set == None:
        print "ERROR : Cannot open dataset : ", "data_set"
        sys.exit(1)

    regionCat = w.obj("channelCat")
    data_set.table(regionCat).Print("v")

    regionFullName = Util.GetFullRegionName(regionCat, region)

    fitRegionsList = fitregions.split(",")
    fitRegionsFullName = ""
    for reg in fitRegionsList:
        regFullName = Util.GetFullRegionName(regionCat, reg)
        if fitRegionsFullName == "":
            fitRegionsFullName = regFullName.Data()
        else:
            fitRegionsFullName = fitRegionsFullName + "," + regFullName.Data()

    chosenSample = False
    if sample is not '':
        chosenSample = True

    #####################################################

    regSys = {}

    regionCatStr = 'channelCat==channelCat::' + regionFullName.Data()
    dataRegion = data_set.reduce(regionCatStr)
    nobsRegion = 0.

    if dataRegion:
        nobsRegion = dataRegion.sumEntries()
    else:
        print " ERROR : dataset-category", regionCatStr, " not found"

    if chosenSample:
        regSys['sqrtnobsa'] = 0.
    else:
        regSys['sqrtnobsa'] = TMath.Sqrt(nobsRegion)

    ####

    if chosenSample:
        pdfInRegion = Util.GetComponent(w, sample, region)
    else:
        rawPdfInRegion = Util.GetRegionPdf(w, region)
        varInRegion = Util.GetRegionVar(w, region)
        prodList = rawPdfInRegion.pdfList()
        foundRRS = 0
        for idx in range(prodList.getSize()):
            if prodList[idx].InheritsFrom("RooRealSumPdf"):
                rrspdfInt = prodList[idx].createIntegral(
                    RooArgSet(varInRegion))
                pdfInRegion = rrspdfInt
                foundRRS += 1
        if foundRRS > 1 or foundRRS == 0:
            print " \n\n WARNING: ", pdf.GetName(
            ), " has ", foundRRS, " instances of RooRealSumPdf"
            print pdf.GetName(), " component list:", prodList.Print("v")

    if not pdfInRegion:
        if chosenSample:
            print " \n Warning, could not find pdf in region = ", region, " for sample = ", sample
        else:
            print " \n Warning, could not find pdf in region = ", region

    nFittedInRegion = pdfInRegion.getVal()
    regSys['sqrtnfitted'] = TMath.Sqrt(nFittedInRegion)

    pdfFittedErrInRegion = Util.GetPropagatedError(pdfInRegion, result)
    regSys['totsyserr'] = pdfFittedErrInRegion

    # redo the fit for every parameter being fixed
    lumiConst = True
    fpf = result.floatParsFinal()

    # redo the fit for every parameter being fixed
    for idx in range(fpf.getSize()):

        parname = fpf[idx].GetName()
        print "\n Method-2: redoing fit with fixed parameter ", parname

        # the parameter that is fixed, needs to have the value of the default fit
        w.loadSnapshot(snapshot)
        par = w.var(parname)

        #     # before redoing the fit, set the values of parameters to initial snapshot, otherwise MIGRAD cannot find improvement
        #     w.loadSnapshot('snapshot_paramsVals_initial')
        #     par.setVal(parDefVal)
        par.setConstant(True)
        suffix = parname + "Fixed"
        result_1parfixed = Util.FitPdf(w, fitRegionsFullName, lumiConst,
                                       data_set, suffix)

        expResultAfter_1parfixed = RooExpandedFitResult(
            result_1parfixed, resultlistOrig)

        nFittedInRegion_1parfixed = pdfInRegion.getVal()
        pdfFittedErrInRegion_1parfixed = Util.GetPropagatedError(
            pdfInRegion, expResultAfter_1parfixed)  #  result_1parfixed)

        if pdfFittedErrInRegion_1parfixed > pdfFittedErrInRegion:
            print "\n\n  WARNING  parameter ", parname, " gives a larger error when set constant. Do you expect this?"
            print "  WARNING          pdfFittedErrInRegion = ", pdfFittedErrInRegion, "    pdfFittedErrInRegion_1parfixed = ", pdfFittedErrInRegion_1parfixed

        systError = TMath.Sqrt(
            abs(pdfFittedErrInRegion * pdfFittedErrInRegion -
                pdfFittedErrInRegion_1parfixed *
                pdfFittedErrInRegion_1parfixed))
        par.setConstant(False)

        if result_1parfixed.status() == 0 and result_1parfixed.covQual(
        ) == 3:  #and result_1parfixed.numStatusHistory()==2 and  result_1parfixed.statusCodeHistory(0)==0 and  result_1parfixed.statusCodeHistory(1) ==0:
            systError = systError
        else:
            systError = 0.0
            print "        WARNING :   for parameter ", parname, " fixed the fit does not converge, as status=", result_1parfixed.status(
            ), "(converged=0),  and covariance matrix quality=", result_1parfixed.covQual(
            ), " (full accurate==3)"
            print "        WARNING: setting systError = 0 for parameter ", parname

        #if namemap.has_key(parname):
        #  parname = namemap[parname]
        regSys['syserr_' + parname] = systError

    return regSys
Exemple #2
0
def latexfitresults(filename,
                    regionList,
                    sampleList,
                    dataname='obsData',
                    showSum=False,
                    doAsym=True,
                    blinded=False,
                    splitBins=False):
    """
  Calculate before/after-fit yields in all channels given
  
  @param filename The filename containing afterFit workspace
  @param regionList A list of regions to be considered
  @param sampleList A list of samples to be considered
  @param dataname The name of dataset (default='obsData')
  @param showSum Calculates sum of all regions if set to true (default=False)
  @param doAsym Calculates asymmetric errors taken from MINOS (default=True)
  @param blinded Observed event count will not be shown if set to True (default=False)
  @param splitBins Calculates bin-by-bin yields for all regions if set to True (default=False)
  """
    """
  pick up workspace from file
  """
    workspacename = 'w'
    w = Util.GetWorkspaceFromFile(filename, 'w')
    if w == None:
        print "ERROR : Cannot open workspace : ", workspacename
        sys.exit(1)
    """
  pick up after-fit RooExpandedFitResult from workspace
  """
    resultAfterFit = w.obj('RooExpandedFitResult_afterFit')
    if resultAfterFit == None:
        print "ERROR : Cannot open fit result after fit RooExpandedFitResult_afterFit"
        sys.exit(1)
    """
  pick up before-fit RooExpandedFitResult from workspace
  """
    resultBeforeFit = w.obj('RooExpandedFitResult_beforeFit')
    if resultBeforeFit == None:
        print "ERROR : Cannot open fit result before fit RooExpandedFitResult_beforeFit"
        sys.exit(1)
    """
  pick up dataset from workspace
  """
    data_set = w.data(dataname)
    if data_set == None:
        print "ERROR : Cannot open dataset : ", "data_set" + suffix
        sys.exit(1)
    """
  pick up channel category (RooCategory) from workspace
  """
    regionCat = w.obj("channelCat")
    if not blinded:
        data_set.table(regionCat).Print("v")
    """
  find full (long) name list of regions (i.e. short=SR3J, long=SR3J_meffInc30_JVF25pt50)
  """
    regionFullNameList = [
        Util.GetFullRegionName(regionCat, region) for region in regionList
    ]
    """
  load afterFit workspace snapshot (=set all parameters to values after fit)
  """
    snapshot = 'snapshot_paramsVals_RooExpandedFitResult_afterFit'
    w.loadSnapshot(snapshot)

    if not w.loadSnapshot(snapshot):
        print "ERROR : Cannot load snapshot : ", snapshot
        sys.exit(1)
    """
  define set, for all names/yields to be saved in
  """
    tablenumbers = {}
    """
  if showSum=True define names for sum of all regions and add to regionList
  """
    sumName = ""
    for index, reg in enumerate(regionList):
        if index == 0:
            sumName = reg
        else:
            sumName = sumName + " + " + reg

    regionListWithSum = list(regionList)
    if showSum:
        regionListWithSum.append(sumName)

    tablenumbers['names'] = regionListWithSum
    """
  make a list of channelCat calls for every region
  """
    regionCatList = [
        'channelCat==channelCat::' + region.Data()
        for region in regionFullNameList
    ]
    """
  retrieve number of observed (=data) events per region
  """
    regionDatasetList = [
        data_set.reduce(regioncat) for regioncat in regionCatList
    ]
    for index, data in enumerate(regionDatasetList):
        data.SetName("data_" + regionList[index])
        data.SetTitle("data_" + regionList[index])

    nobs_regionList = [data.sumEntries() for data in regionDatasetList]
    """
  if showSum=True calculate the total number of observed events in all regions  
  """
    sumNobs = 0.
    for nobs in nobs_regionList:
        sumNobs += nobs
    if showSum:
        nobs_regionList.append(sumNobs)
    tablenumbers['nobs'] = nobs_regionList
    """
  FROM HERE ON OUT WE CALCULATE THE FITTED NUMBER OF EVENTS __AFTER__ THE FIT
  """
    """
  get a list of pdf's and variables per region
  """
    pdfinRegionList = [Util.GetRegionPdf(w, region) for region in regionList]
    varinRegionList = [Util.GetRegionVar(w, region) for region in regionList]
    """
  if splitBins=True get the list of Nbins, binMax and binMin; make a list of new region names for each bin
  """
    varNbinsInRegionList = []
    varBinLowInRegionList = []
    varBinHighInRegionList = []
    rangeNameBinsInRegionList = []
    if splitBins:
        varNbinsInRegionList = [
            Util.GetRegionVar(w, region).getBinning().numBins()
            for region in regionList
        ]
        varBinLowInRegionList = [[
            Util.GetRegionVar(w, region).getBinning(
                (region + "binning")).binLow(ibin)
            for ibin in range(0, varNbinsInRegionList[idx])
        ] for idx, region in enumerate(regionList)]
        varBinHighInRegionList = [[
            Util.GetRegionVar(w, region).getBinning(
                (region + "binning")).binHigh(ibin)
            for ibin in range(0, varNbinsInRegionList[idx])
        ] for idx, region in enumerate(regionList)]
        rangeNameBinsInRegionList = [[
            regionList[idx] + "_bin" + str(ibin)
            for ibin in range(0, varNbinsInRegionList[idx])
        ] for idx, region in enumerate(regionList)]
        for index, region in enumerate(regionList):
            if varNbinsInRegionList[index] == 1:
                print " \n YieldsTable.py: WARNING: you have called -P (= per-bin yields) but this region ", region, " has only 1 bin \n"
    """
  if splitBins=True reshuffle the regionName list; each region name is followed by names of each bin (i.e. regionNameList=['SR3J','SR3J_bin1','SR3j_bin2','SR4J','SR4J_bin1'])
  """
    regionListWithBins = []
    if splitBins:
        for index, region in enumerate(regionList):
            regionListWithBins.append(region)
            for ibin in range(0, varNbinsInRegionList[index]):
                regionListWithBins.append(
                    rangeNameBinsInRegionList[index][ibin])
        tablenumbers['names'] = regionListWithBins
    """
  calculate number of observed(=data) events per bin
  """
    nobs_regionListWithBins = []
    if splitBins:
        binFuncInRegionList = [
            RooBinningCategory("bin_" + region, "bin_" + region,
                               varinRegionList[index])
            for index, region in enumerate(regionList)
        ]
        for index, data in enumerate(regionDatasetList):
            data.addColumn(binFuncInRegionList[index])
            if not blinded:
                data.table(binFuncInRegionList[index]).Print("v")
            nobs_regionListWithBins.append(data.sumEntries())
            for ibin in range(0, varNbinsInRegionList[index]):
                nobs_regionListWithBins.append(
                    (data.reduce(binFuncInRegionList[index].GetName() + "==" +
                                 binFuncInRegionList[index].GetName() + "::" +
                                 varinRegionList[index].GetName() + "_bin" +
                                 str(ibin))).sumEntries())

        tablenumbers['nobs'] = nobs_regionListWithBins
    """
  if blinded=True, set all numbers of observed events to -1
  """
    if blinded:
        for index, nobs in enumerate(nobs_regionListWithBins):
            nobs_regionListWithBins[index] = -1
        tablenumbers['nobs'] = nobs_regionListWithBins
    """
  get a list of RooRealSumPdf per region (RooRealSumPdf is the top-pdf per region containing all samples)
  """
    rrspdfinRegionList = []
    for index, pdf in enumerate(pdfinRegionList):
        prodList = pdf.pdfList()
        foundRRS = 0
        for idx in range(prodList.getSize()):
            if prodList[idx].InheritsFrom("RooRealSumPdf"):
                rrspdfInt = prodList[idx].createIntegral(
                    RooArgSet(varinRegionList[index]))
                rrspdfinRegionList.append(rrspdfInt)
                if splitBins:
                    origMin = varinRegionList[index].getMin()
                    origMax = varinRegionList[index].getMax()
                    for ibin in range(0, varNbinsInRegionList[index]):
                        rangeName = rangeNameBinsInRegionList[index][ibin]
                        varinRegionList[index].setRange(
                            rangeName, varBinLowInRegionList[index][ibin],
                            varBinHighInRegionList[index][ibin])
                        rrspdfInt = prodList[idx].createIntegral(
                            RooArgSet(varinRegionList[index]), rangeName)
                        rrspdfinRegionList.append(rrspdfInt)
                    varinRegionList[index].setRange(origMin, origMax)
                foundRRS += 1
        if foundRRS > 1 or foundRRS == 0:
            print " \n\n WARNING: ", pdf.GetName(
            ), " has ", foundRRS, " instances of RooRealSumPdf"
            print pdf.GetName(), " component list:", prodList.Print("v")
    """
  calculate total pdf number of fitted events and error
  """
    nFittedInRegionList = [
        pdf.getVal() for index, pdf in enumerate(rrspdfinRegionList)
    ]
    pdfFittedErrInRegionList = [
        Util.GetPropagatedError(pdf, resultAfterFit, doAsym)
        for pdf in rrspdfinRegionList
    ]
    """
  if showSum=True calculate the total number of fitted events in all regions  
  """
    if showSum:
        pdfInAllRegions = RooArgSet()
        for index, pdf in enumerate(rrspdfinRegionList):
            pdfInAllRegions.add(pdf)
        pdfSumInAllRegions = RooAddition("pdf_AllRegions_AFTER",
                                         "pdf_AllRegions_AFTER",
                                         RooArgList(pdfInAllRegions))
        nPdfSumVal = pdfSumInAllRegions.getVal()
        nPdfSumError = Util.GetPropagatedError(pdfSumInAllRegions,
                                               resultAfterFit, doAsym)
        nFittedInRegionList.append(nPdfSumVal)
        pdfFittedErrInRegionList.append(nPdfSumError)

    tablenumbers['TOTAL_FITTED_bkg_events'] = nFittedInRegionList
    tablenumbers['TOTAL_FITTED_bkg_events_err'] = pdfFittedErrInRegionList
    """
  calculate the fitted number of events and propagated error for each requested sample, by splitting off each sample pdf
  """
    for isam, sample in enumerate(sampleList):
        sampleName = getName(sample)
        nSampleInRegionVal = []
        nSampleInRegionError = []
        sampleInAllRegions = RooArgSet()
        for ireg, region in enumerate(regionList):
            sampleInRegion = getPdfInRegions(w, sample, region)
            sampleInRegionVal = 0.
            sampleInRegionError = 0.
            if not sampleInRegion == None:
                sampleInRegionVal = sampleInRegion.getVal()
                sampleInRegionError = Util.GetPropagatedError(
                    sampleInRegion, resultAfterFit, doAsym)
                sampleInAllRegions.add(sampleInRegion)
            else:
                print " \n YieldsTable.py: WARNING: sample =", sampleName, " non-existent (empty) in region =", region, "\n"
            nSampleInRegionVal.append(sampleInRegionVal)
            nSampleInRegionError.append(sampleInRegionError)
            """
      if splitBins=True calculate numbers of fitted events plus error per bin      
      """
            if splitBins:
                origMin = varinRegionList[ireg].getMin()
                origMax = varinRegionList[ireg].getMax()
                for ibin in range(0, varNbinsInRegionList[ireg]):
                    rangeName = rangeNameBinsInRegionList[ireg][ibin]
                    sampleInRegion = getPdfInRegionsWithRangeName(
                        w, sample, region, rangeName)
                    sampleInRegionVal = 0.
                    sampleInRegionError = 0.
                    if not sampleInRegion == None:
                        varinRegionList[ireg].setRange(
                            rangeName, varBinLowInRegionList[ireg][ibin],
                            varBinHighInRegionList[ireg][ibin])
                        sampleInRegionVal = sampleInRegion.getVal()
                        sampleInRegionError = Util.GetPropagatedError(
                            sampleInRegion, resultAfterFit, doAsym)
                    else:
                        print " \n YieldsTable.py: WARNING: sample =", sampleName, " non-existent (empty) in region=", region, " bin=", ibin, " \n"
                    nSampleInRegionVal.append(sampleInRegionVal)
                    nSampleInRegionError.append(sampleInRegionError)

                varinRegionList[ireg].setRange(origMin, origMax)
        """
    if showSum=True calculate the total number of fitted events in all regions  
    """
        if showSum:
            sampleSumInAllRegions = RooAddition(
                (sampleName + "_AllRegions_FITTED"),
                (sampleName + "_AllRegions_FITTED"),
                RooArgList(sampleInAllRegions))
            nSampleSumVal = sampleSumInAllRegions.getVal()
            nSampleSumError = Util.GetPropagatedError(sampleSumInAllRegions,
                                                      resultAfterFit, doAsym)
            nSampleInRegionVal.append(nSampleSumVal)
            nSampleInRegionError.append(nSampleSumError)
        tablenumbers['Fitted_events_' + sampleName] = nSampleInRegionVal
        tablenumbers['Fitted_err_' + sampleName] = nSampleInRegionError

    print "\n starting BEFORE-FIT calculations \n"
    """
  FROM HERE ON OUT WE CALCULATE THE EXPECTED NUMBER OF EVENTS __BEFORRE__ THE FIT
  """
    """
  load beforeFit workspace snapshot (=set all parameters to values before fit)
  """
    w.loadSnapshot('snapshot_paramsVals_RooExpandedFitResult_beforeFit')
    """
  check if any of the initial scaling factors is != 1
  """
    _result = w.obj('RooExpandedFitResult_beforeFit')
    _muFacs = _result.floatParsFinal()

    for i in range(len(_muFacs)):
        if "mu_" in _muFacs[i].GetName() and _muFacs[i].getVal() != 1.0:
            print " \n WARNING: scaling factor %s != 1.0 (%g) expected MC yield WILL BE WRONG!" % (
                _muFacs[i].GetName(), _muFacs[i].getVal())
    """
  get a list of pdf's and variables per region
  """
    pdfinRegionList = [Util.GetRegionPdf(w, region) for region in regionList]
    varinRegionList = [Util.GetRegionVar(w, region) for region in regionList]
    """
  get a list of RooRealSumPdf per region (RooRealSumPdf is the top-pdf per region containing all samples)
  """
    rrspdfinRegionList = []
    for index, pdf in enumerate(pdfinRegionList):
        prodList = pdf.pdfList()
        foundRRS = 0
        for idx in range(prodList.getSize()):
            if prodList[idx].InheritsFrom("RooRealSumPdf"):
                rrspdfInt = prodList[idx].createIntegral(
                    RooArgSet(varinRegionList[index]))
                rrspdfinRegionList.append(rrspdfInt)
                if splitBins:
                    origMin = varinRegionList[index].getMin()
                    origMax = varinRegionList[index].getMax()
                    for ibin in range(0, varNbinsInRegionList[index]):
                        rangeName = rangeNameBinsInRegionList[index][ibin]
                        varinRegionList[index].setRange(
                            rangeName, varBinLowInRegionList[index][ibin],
                            varBinHighInRegionList[index][ibin])
                        rrspdfInt = prodList[idx].createIntegral(
                            RooArgSet(varinRegionList[index]), rangeName)
                        rrspdfinRegionList.append(rrspdfInt)
                    varinRegionList[index].setRange(origMin, origMax)
                foundRRS += 1
        if foundRRS > 1 or foundRRS == 0:
            print " \n\n WARNING: ", pdf.GetName(
            ), " has ", foundRRS, " instances of RooRealSumPdf"
            print pdf.GetName(), " component list:", prodList.Print("v")
    """
  calculate total pdf number of expected events and error
  """
    nExpInRegionList = [
        pdf.getVal() for index, pdf in enumerate(rrspdfinRegionList)
    ]
    pdfExpErrInRegionList = [
        Util.GetPropagatedError(pdf, resultBeforeFit, doAsym)
        for pdf in rrspdfinRegionList
    ]
    """
  if showSum=True calculate the total number of expected events in all regions  
  """
    if showSum:
        pdfInAllRegions = RooArgSet()
        for index, pdf in enumerate(rrspdfinRegionList):
            pdfInAllRegions.add(pdf)
        pdfSumInAllRegions = RooAddition("pdf_AllRegions_BEFORE",
                                         "pdf_AllRegions_BEFORE",
                                         RooArgList(pdfInAllRegions))
        nPdfSumVal = pdfSumInAllRegions.getVal()
        nPdfSumError = Util.GetPropagatedError(pdfSumInAllRegions,
                                               resultBeforeFit, doAsym)
        nExpInRegionList.append(nPdfSumVal)
        pdfExpErrInRegionList.append(nPdfSumError)

    tablenumbers['TOTAL_MC_EXP_BKG_events'] = nExpInRegionList
    tablenumbers['TOTAL_MC_EXP_BKG_err'] = pdfExpErrInRegionList
    """
  calculate the fitted number of events and propagated error for each requested sample, by splitting off each sample pdf
  """
    for isam, sample in enumerate(sampleList):
        sampleName = getName(sample)
        nMCSampleInRegionVal = []
        nMCSampleInRegionError = []
        MCSampleInAllRegions = RooArgSet()
        for ireg, region in enumerate(regionList):
            MCSampleInRegion = getPdfInRegions(w, sample, region)
            MCSampleInRegionVal = 0.
            MCSampleInRegionError = 0.
            if not MCSampleInRegion == None:
                MCSampleInRegionVal = MCSampleInRegion.getVal()
                MCSampleInRegionError = Util.GetPropagatedError(
                    MCSampleInRegion, resultBeforeFit, doAsym)
                MCSampleInAllRegions.add(MCSampleInRegion)
            else:
                print " \n WARNING: sample=", sampleName, " non-existent (empty) in region=", region
            nMCSampleInRegionVal.append(MCSampleInRegionVal)
            nMCSampleInRegionError.append(MCSampleInRegionError)
            """
      if splitBins=True calculate numbers of fitted events plus error per bin      
      """
            if splitBins:
                origMin = varinRegionList[ireg].getMin()
                origMax = varinRegionList[ireg].getMax()
                for ibin in range(0, varNbinsInRegionList[ireg]):
                    rangeName = rangeNameBinsInRegionList[ireg][ibin]
                    MCSampleInRegion = getPdfInRegionsWithRangeName(
                        w, sample, region, rangeName)
                    MCSampleInRegionVal = 0.
                    MCSampleInRegionError = 0.
                    if not MCSampleInRegion == None:
                        varinRegionList[ireg].setRange(
                            rangeName, varBinLowInRegionList[ireg][ibin],
                            varBinHighInRegionList[ireg][ibin])
                        MCSampleInRegionVal = MCSampleInRegion.getVal()
                        MCSampleInRegionError = Util.GetPropagatedError(
                            MCSampleInRegion, resultBeforeFit, doAsym)
                    else:
                        print " \n YieldsTable.py: WARNING: sample =", sampleName, " non-existent (empty) in region=", region, " bin=", ibin, " \n"
                    nMCSampleInRegionVal.append(MCSampleInRegionVal)
                    nMCSampleInRegionError.append(MCSampleInRegionError)

                varinRegionList[ireg].setRange(origMin, origMax)
        """
    if showSum=True calculate the total number of fitted events in all regions  
    """
        if showSum:
            MCSampleSumInAllRegions = RooAddition(
                (sampleName + "_AllRegions_MC"),
                (sampleName + "_AllRegions_MC"),
                RooArgList(MCSampleInAllRegions))
            nMCSampleSumVal = MCSampleSumInAllRegions.getVal()
            nMCSampleSumError = Util.GetPropagatedError(
                MCSampleSumInAllRegions, resultBeforeFit, doAsym)
            nMCSampleInRegionVal.append(nMCSampleSumVal)
            nMCSampleInRegionError.append(nMCSampleSumError)
        tablenumbers['MC_exp_events_' + sampleName] = nMCSampleInRegionVal
        tablenumbers['MC_exp_err_' + sampleName] = nMCSampleInRegionError
    """
  sort the tablenumbers set
  """
    map_listofkeys = tablenumbers.keys()
    map_listofkeys.sort()
    """
  print the sorted tablenumbers set
  """
    for name in map_listofkeys:
        if tablenumbers.has_key(name):
            print name, ": ", tablenumbers[name]

    return tablenumbers
def latexfitresults_method2(filename,
                            resultname='RooExpandedFitResult_afterFit',
                            region='3jL',
                            sample='',
                            fitregions='WR,TR,S3,S4,SR3jT,SR4jT',
                            dataname='obsData',
                            doAsym=False,
                            SRName=""):
    """
  Method-2: set the parameter you're interested in constant,
  redo the fit with all other parameters floating,
  calculate the quadratic difference between default fit and your new model with parameter fixed

  @param filename The filename containing afterFit workspace
  @param resultname The name of fit result (typically='RooExpandedFitResult_afterFit' or 'RooExpandedFitResult_beforeFit'
  @param region The region to be used for systematics breakdown calculation
  @param sample The sample to be used insted of total pdf (default='' not defined, hence total pdf used)
  @param fitregions Fit regions to perform the re-fit (default= 'WR,TR,S3,S4,SR3jT,SR4jT' but needs to be specified by user)
  @param dataname The name of dataset (default='obsData')
  @param doAsym Calculates asymmetric errors taken from MINOS (default=False) 
  """
    """
  pick up workspace from file
  """
    w = Util.GetWorkspaceFromFile(filename, 'w')
    if w == None:
        print "ERROR : Cannot open workspace : "
        sys.exit(1)
    """
  pick up RooExpandedFitResult from workspace with name resultName (either before or after fit)
  """
    result = w.obj(resultname)
    if result == None:
        print "ERROR : Cannot open fit result : ", resultname
        sys.exit(1)
    """
  save the original (after-fit result) fit parameters list
  """
    resultlistOrig = result.floatParsFinal()
    """
  load workspace snapshot related to resultName (=set all parameters to values after fit)
  """
    snapshot = 'snapshot_paramsVals_' + resultname
    w.loadSnapshot(snapshot)
    """
  pick up dataset from workspace
  """
    data_set = w.data(dataname)
    if data_set == None:
        print "ERROR : Cannot open dataset : ", "data_set"
        sys.exit(1)
    """
  pick up channel category (RooCategory) from workspace
  """
    regionCat = w.obj("channelCat")
    #  data_set.table(regionCat).Print("v");
    """
  find full (long) name list of region (i.e. short=SR3J, long=SR3J_meffInc30_JVF25pt50)
  """
    regionFullName = Util.GetFullRegionName(regionCat, region)
    """
  find and save the list of all regions used for the fit, as the fit will be redone
  """
    fitRegionsList = fitregions.split(",")
    fitRegionsFullName = ""
    for reg in fitRegionsList:
        regFullName = Util.GetFullRegionName(regionCat, reg)
        if fitRegionsFullName == "":
            fitRegionsFullName = regFullName.Data()
        else:
            fitRegionsFullName = fitRegionsFullName + "," + regFullName.Data()
    """
  set a boolean whether we're looking at a sample or the full (multi-sample) pdf/model
  """
    chosenSample = False
    if sample is not '':
        chosenSample = True
    """
  define regSys set, for all names/numbers to be saved in
  """
    regSys = {}
    """
  define channelCat call for this region and reduce the dataset to this category/region
  """
    regionCatStr = 'channelCat==channelCat::' + regionFullName.Data()
    dataRegion = data_set.reduce(regionCatStr)
    nobsRegion = 0.

    if dataRegion:
        nobsRegion = dataRegion.sumEntries()
    else:
        print " ERROR : dataset-category", regionCatStr, " not found"
    """
  if looking at a sample, there is no equivalent N_obs (only for the full model)
  """
    if chosenSample:
        regSys['sqrtnobsa'] = 0.
    else:
        regSys['sqrtnobsa'] = TMath.Sqrt(nobsRegion)
    """
  get the pdf for the total model or just for the sample in region
  """
    if chosenSample:
        pdfInRegion = Util.GetComponent(w, sample, region)
    else:
        rawPdfInRegion = Util.GetRegionPdf(w, region)
        varInRegion = Util.GetRegionVar(w, region)
        prodList = rawPdfInRegion.pdfList()
        foundRRS = 0
        for idx in range(prodList.getSize()):
            if prodList[idx].InheritsFrom("RooRealSumPdf"):
                rrspdfInt = prodList[idx].createIntegral(
                    RooArgSet(varInRegion))
                pdfInRegion = rrspdfInt
                foundRRS += 1
        if foundRRS > 1 or foundRRS == 0:
            print " \n\n WARNING: ", pdf.GetName(
            ), " has ", foundRRS, " instances of RooRealSumPdf"
            print pdf.GetName(), " component list:", prodList.Print("v")

    if not pdfInRegion:
        if chosenSample:
            print " \n Warning, could not find pdf in region = ", region, " for sample = ", sample
        else:
            print " \n Warning, could not find pdf in region = ", region
    """
  calculate fitted pdf number of events and full error
  """
    nFittedInRegion = pdfInRegion.getVal()
    regSys['sqrtnfitted'] = TMath.Sqrt(nFittedInRegion)
    regSys['nfitted'] = nFittedInRegion

    pdfFittedErrInRegion = Util.GetPropagatedError(pdfInRegion, result, doAsym)
    regSys['totsyserr'] = pdfFittedErrInRegion
    """
  set lumi parameter constant for the refit -- FIXME
  """
    lumiConst = True

    fpf = result.floatParsFinal()
    """
  redo the fit for every parameter being fixed
  """
    for idx in range(fpf.getSize()):
        parname = fpf[idx].GetName()
        print "\n Method-2: redoing fit with fixed parameter ", parname
        """
    the parameter that is fixed, needs to have the value of the default fit
    """
        w.loadSnapshot(snapshot)
        par = w.var(parname)
        par.setConstant(True)
        """
    perform the fit again with one parameter fixed
    """
        suffix = parname + "Fixed"
        result_1parfixed = Util.FitPdf(w, fitRegionsFullName, lumiConst,
                                       data_set, suffix, doAsym, "all")
        """
    create a new RooExpandedFitResult based on the new fit
     and all parameters saved in the original fit result (as some parameters might only be floating in VRs)
    """
        expResultAfter_1parfixed = RooExpandedFitResult(
            result_1parfixed, resultlistOrig)
        """
    calculate newly fitted number of events and full error
    """
        nFittedInRegion_1parfixed = pdfInRegion.getVal()
        pdfFittedErrInRegion_1parfixed = Util.GetPropagatedError(
            pdfInRegion, expResultAfter_1parfixed,
            doAsym)  #  result_1parfixed)
        """
    check whether original total error is smaller then newly-fitted total error
      if one does anew fit with less floating parameters (systematics), it can be expected to see smaller error
      (this assumption does not take correlations into account)
    """
        if pdfFittedErrInRegion_1parfixed > pdfFittedErrInRegion:
            print "\n\n  WARNING  parameter ", parname, " gives a larger error when set constant. Do you expect this?"
            print "  WARNING          pdfFittedErrInRegion = ", pdfFittedErrInRegion, "    pdfFittedErrInRegion_1parfixed = ", pdfFittedErrInRegion_1parfixed
        """
    calculate systematic error as the quadratic difference between original and re-fitted errors
    """
        systError = TMath.Sqrt(
            abs(pdfFittedErrInRegion * pdfFittedErrInRegion -
                pdfFittedErrInRegion_1parfixed *
                pdfFittedErrInRegion_1parfixed))
        par.setConstant(False)
        """
    print a warning if new fit with 1 par fixed did not converge - meaning that sys error cannot be trusted 
    """
        if result_1parfixed.status() == 0 and result_1parfixed.covQual(
        ) == 3:  #and result_1parfixed.numStatusHistory()==2 and  result_1parfixed.statusCodeHistory(0)==0 and  result_1parfixed.statusCodeHistory(1) ==0:
            systError = systError
        else:
            systError = 0.0
            print "        WARNING :   for parameter ", parname, " fixed the fit does not converge, as status=", result_1parfixed.status(
            ), "(converged=0),  and covariance matrix quality=", result_1parfixed.covQual(
            ), " (full accurate==3)"
            print "        WARNING: setting systError = 0 for parameter ", parname

        regSys['syserr_' + parname] = systError

    return regSys
Exemple #4
0
def latexfitresults(filename,
                    region='3jL',
                    sample='',
                    resultName="RooExpandedFitResult_afterFit",
                    dataname='obsData'):

    #namemap = {}
    #namemap = getnamemap()

    ############################################

    w = Util.GetWorkspaceFromFile(filename, 'w')

    if w == None:
        print "ERROR : Cannot open workspace : ", workspacename
        sys.exit(1)

    result = w.obj(resultName)
    if result == None:
        print "ERROR : Cannot open fit result ", resultName
        sys.exit(1)

    snapshot = 'snapshot_paramsVals_' + resultName
    w.loadSnapshot(snapshot)

    data_set = w.data(dataname)
    if data_set == None:
        print "ERROR : Cannot open dataset : ", "data_set"
        sys.exit(1)

    regionCat = w.obj("channelCat")
    data_set.table(regionCat).Print("v")

    regionFullName = Util.GetFullRegionName(regionCat, region)

    chosenSample = False
    if sample is not '':
        chosenSample = True

    #####################################################

    regSys = {}

    regionCatStr = 'channelCat==channelCat::' + regionFullName.Data()
    dataRegion = data_set.reduce(regionCatStr)

    nobsRegion = 0.

    if dataRegion:
        nobsRegion = dataRegion.sumEntries()
    else:
        print " ERROR : dataset-category dataRegion not found"

    if chosenSample:
        regSys['sqrtnobsa'] = 0.
    else:
        regSys['sqrtnobsa'] = TMath.Sqrt(nobsRegion)

    ####

    if chosenSample:
        pdfInRegion = Util.GetComponent(w, sample, region)
    else:
        rawPdfInRegion = Util.GetRegionPdf(w, region)
        varInRegion = Util.GetRegionVar(w, region)
        prodList = rawPdfInRegion.pdfList()
        foundRRS = 0
        for idx in range(prodList.getSize()):
            if prodList[idx].InheritsFrom("RooRealSumPdf"):
                rrspdfInt = prodList[idx].createIntegral(
                    RooArgSet(varInRegion))
                pdfInRegion = rrspdfInt
                foundRRS += 1
        if foundRRS > 1 or foundRRS == 0:
            print " \n\n WARNING: ", pdf.GetName(
            ), " has ", foundRRS, " instances of RooRealSumPdf"
            print pdf.GetName(), " component list:", prodList.Print("v")

    if not pdfInRegion:
        if chosenSample:
            print " \n Warning, could not find pdf in region = ", region, " for sample = ", sample
        else:
            print " \n Warning, could not find pdf in region = ", region

    nFittedInRegion = pdfInRegion.getVal()
    regSys['sqrtnfitted'] = TMath.Sqrt(nFittedInRegion)

    pdfFittedErrInRegion = Util.GetPropagatedError(pdfInRegion, result)
    regSys['totsyserr'] = pdfFittedErrInRegion

    # calculate error per parameter on  fitresult
    fpf = result.floatParsFinal()

    # set all floating parameters constant
    for idx in range(fpf.getSize()):
        parname = fpf[idx].GetName()
        par = w.var(parname)
        par.setConstant()

    for idx in range(fpf.getSize()):
        parname = fpf[idx].GetName()
        par = w.var(parname)
        par.setConstant(False)
        sysError = Util.GetPropagatedError(pdfInRegion, result)
        #if namemap.has_key(parname):
        #  parname = namemap[parname]
        regSys['syserr_' + parname] = sysError
        par.setConstant()

    return regSys
def latexfitresults(filename,
                    namemap,
                    region='3jL',
                    sample='',
                    resultName="RooExpandedFitResult_afterFit",
                    dataname='obsData',
                    doAsym=True,
                    SRName=""):
    """
  Method-1: set all parameters constant, except for the one you're interested in, 
           calculate the systematic/error propagated due to that parameter

  @param filename The filename containing afterFit workspace
  @param namemap Defines whether any systematics need to be grouped in calculation (by default not defined, hence each parameter gets used one by one)
  @param resultname The name of fit result (typically='RooExpandedFitResult_afterFit' or 'RooExpandedFitResult_beforeFit'
  @param region The region to be used for systematics breakdown calculation
  @param sample The sample to be used insted of total pdf (default='' not defined, hence total pdf used)
  @param dataname The name of dataset (default='obsData')
  @param doAsym Calculates asymmetric errors taken from MINOS (default=True)
"""
    """
  pick up workspace from file
  """
    workspacename = 'w'
    w = Util.GetWorkspaceFromFile(filename, workspacename)
    if w == None:
        print "ERROR : Cannot open workspace : ", workspacename
        sys.exit(1)
    """
  pick up RooExpandedFitResult from workspace with name resultName (either before or after fit)
  """
    result = w.obj(resultName)
    if result == None:
        print "ERROR : Cannot open fit result ", resultName
        sys.exit(1)
    """
  load workspace snapshot related to resultName (=set all parameters to values after fit)
  """
    snapshot = 'snapshot_paramsVals_' + resultName
    w.loadSnapshot(snapshot)
    """
  pick up dataset from workspace
  """
    data_set = w.data(dataname)
    if data_set == None:
        print "ERROR : Cannot open dataset : ", "data_set"
        sys.exit(1)
    """
  pick up channel category (RooCategory) from workspace
  """
    regionCat = w.obj("channelCat")
    data_set.table(regionCat).Print("v")
    """
  find full (long) name list of region (i.e. short=SR3J, long=SR3J_meffInc30_JVF25pt50)
  """
    regionFullName = Util.GetFullRegionName(regionCat, region)
    """
  set a boolean whether we're looking at a sample or the full (multi-sample) pdf/model
  """
    chosenSample = False
    if sample is not '':
        chosenSample = True
    """
  define regSys set, for all names/numbers to be saved in
  """
    regSys = {}
    """
  define channelCat call for this region and reduce the dataset to this category/region
  """
    regionCatStr = 'channelCat==channelCat::' + regionFullName.Data()
    dataRegion = data_set.reduce(regionCatStr)
    """
  retrieve and save number of observed (=data) events in region
  """
    nobsRegion = 0.
    if dataRegion:
        nobsRegion = dataRegion.sumEntries()
    else:
        print " ERROR : dataset-category dataRegion not found"
    """
  if looking at a sample, there is no equivalent N_obs (only for the full model)
  """
    if chosenSample:
        regSys['sqrtnobsa'] = 0.
    else:
        regSys['sqrtnobsa'] = TMath.Sqrt(nobsRegion)
    """
  get the pdf for the total model or just for the sample in region
  """
    if chosenSample:
        pdfInRegion = getPdfInRegions(w, sample, region)
    else:
        rawPdfInRegion = Util.GetRegionPdf(w, region)
        varInRegion = Util.GetRegionVar(w, region)
        prodList = rawPdfInRegion.pdfList()
        foundRRS = 0
        for idx in range(prodList.getSize()):
            if prodList[idx].InheritsFrom("RooRealSumPdf"):
                rrspdfInt = prodList[idx].createIntegral(
                    RooArgSet(varInRegion))
                pdfInRegion = rrspdfInt
                foundRRS += 1
        if foundRRS > 1 or foundRRS == 0:
            print " \n\n WARNING: ", pdf.GetName(
            ), " has ", foundRRS, " instances of RooRealSumPdf"
            print pdf.GetName(), " component list:", prodList.Print("v")

    if not pdfInRegion:
        if chosenSample:
            print " \n Warning, could not find pdf in region = ", region, " for sample = ", sample
        else:
            print " \n Warning, could not find pdf in region = ", region
    """
  calculate fitted pdf number of events and full error
  """
    nFittedInRegion = pdfInRegion.getVal()
    regSys['sqrtnfitted'] = TMath.Sqrt(nFittedInRegion)
    regSys['nfitted'] = nFittedInRegion

    pdfFittedErrInRegion = Util.GetPropagatedError(pdfInRegion, result, doAsym)
    regSys['totsyserr'] = pdfFittedErrInRegion
    """
  MakePlots
  """
    makeNicePlots(result, w, SRName)
    """
  calculate error per (floating) parameter in fitresult
  get a list of floating parameters to loop over
  """
    fpf = result.floatParsFinal()
    """
  set all floating parameters constant
  """
    for idx in range(fpf.getSize()):
        parname = fpf[idx].GetName()
        par = w.var(parname)
        par.setConstant()
    """
  if several systematatic/parameters are pre-defined in namemap, they will be floated together
  or in other words, one will get the error due to all pre-defined systematics
  """
    """
  else, float each parameter one by one and calculate the error due to it
  """
    if len(namemap) > 0:
        for key in namemap.keys():
            print namemap[key]
            for parname in namemap[key]:
                par = w.var(parname)
                par.setConstant(False)
                pass
            sysError = Util.GetPropagatedError(pdfInRegion, result, doAsym)
            regSys['syserr_' + key] = sysError
            for idx in range(fpf.getSize()):
                parname = fpf[idx].GetName()
                par = w.var(parname)
                par.setConstant()
                pass
    else:
        for idx in range(fpf.getSize()):
            parname = fpf[idx].GetName()
            par = w.var(parname)
            par.setConstant(False)
            sysError = Util.GetPropagatedError(pdfInRegion, result, doAsym)
            regSys['syserr_' + parname] = sysError
            par.setConstant()
            print parname, par.getVal(), par.getError()

    return regSys
Exemple #6
0
def latexfitresults(filename,
                    regionList,
                    sampleList,
                    dataname='obsData',
                    showSum=False):

    w = Util.GetWorkspaceFromFile(filename, 'w')

    if w == None:
        print "ERROR : Cannot open workspace : ", workspacename
        sys.exit(1)

    resultAfterFit = w.obj('RooExpandedFitResult_afterFit')
    if resultAfterFit == None:
        print "ERROR : Cannot open fit result after fit RooExpandedFitResult_afterFit"
        sys.exit(1)

    resultBeforeFit = w.obj('RooExpandedFitResult_beforeFit')
    if resultBeforeFit == None:
        print "ERROR : Cannot open fit result before fit RooExpandedFitResult_beforeFit"
        sys.exit(1)

    data_set = w.data(dataname)
    if data_set == None:
        print "ERROR : Cannot open dataset : ", "data_set" + suffix
        sys.exit(1)

    regionCat = w.obj("channelCat")
    data_set.table(regionCat).Print("v")

    regionFullNameList = [
        Util.GetFullRegionName(regionCat, region) for region in regionList
    ]
    print regionFullNameList

    ###

    snapshot = 'snapshot_paramsVals_RooExpandedFitResult_afterFit'
    w.loadSnapshot(snapshot)

    if not w.loadSnapshot(snapshot):
        print "ERROR : Cannot load snapshot : ", snapshot
        sys.exit(1)

    tablenumbers = {}

    # SUM ALL REGIONS
    sumName = ""
    for index, reg in enumerate(regionList):
        if index == 0:
            sumName = reg
        else:
            sumName = sumName + " + " + reg

    regionListWithSum = list(regionList)
    if showSum:
        regionListWithSum.append(sumName)

    tablenumbers['names'] = regionListWithSum

    regionCatList = [
        'channelCat==channelCat::' + region.Data()
        for region in regionFullNameList
    ]

    regionDatasetList = [
        data_set.reduce(regioncat) for regioncat in regionCatList
    ]
    for index, data in enumerate(regionDatasetList):
        data.SetName("data_" + regionList[index])
        data.SetTitle("data_" + regionList[index])

    nobs_regionList = [data.sumEntries() for data in regionDatasetList]
    #SUM
    sumNobs = 0.
    for nobs in nobs_regionList:
        sumNobs += nobs
        print " \n XXX nobs = ", nobs, "    sumNobs = ", sumNobs
    if showSum:
        nobs_regionList.append(sumNobs)
    tablenumbers['nobs'] = nobs_regionList

    ######
    ######
    ######  FROM HERE ON OUT WE CALCULATE THE FITTED NUMBER OF EVENTS __AFTER__ THE FIT
    ######
    ######

    # total pdf, not splitting in components
    pdfinRegionList = [Util.GetRegionPdf(w, region) for region in regionList]
    varinRegionList = [Util.GetRegionVar(w, region) for region in regionList]
    rrspdfinRegionList = []
    for index, pdf in enumerate(pdfinRegionList):
        #    pdf.Print("t")
        prodList = pdf.pdfList()
        foundRRS = 0
        for idx in range(prodList.getSize()):
            #      if "BG" in prodList[idx].GetName():
            #        prodList[idx].Print("t")
            if prodList[idx].InheritsFrom("RooRealSumPdf"):
                rrspdfInt = prodList[idx].createIntegral(
                    RooArgSet(varinRegionList[index]))
                rrspdfinRegionList.append(rrspdfInt)
                foundRRS += 1
        if foundRRS > 1 or foundRRS == 0:
            print " \n\n WARNING: ", pdf.GetName(
            ), " has ", foundRRS, " instances of RooRealSumPdf"
            print pdf.GetName(), " component list:", prodList.Print("v")

    nFittedInRegionList = [
        pdf.getVal() for index, pdf in enumerate(rrspdfinRegionList)
    ]
    pdfFittedErrInRegionList = [
        Util.GetPropagatedError(pdf, resultAfterFit)
        for pdf in rrspdfinRegionList
    ]

    if showSum:
        pdfInAllRegions = RooArgSet()
        for index, pdf in enumerate(rrspdfinRegionList):
            pdfInAllRegions.add(pdf)
        pdfSumInAllRegions = RooAddition("pdf_AllRegions_AFTER",
                                         "pdf_AllRegions_AFTER",
                                         pdfInAllRegions)
        pdfSumInAllRegions.Print()
        nPdfSumVal = pdfSumInAllRegions.getVal()
        nPdfSumError = Util.GetPropagatedError(pdfSumInAllRegions,
                                               resultAfterFit)
        nFittedInRegionList.append(nPdfSumVal)
        pdfFittedErrInRegionList.append(nPdfSumError)

    tablenumbers['TOTAL_FITTED_bkg_events'] = nFittedInRegionList
    tablenumbers['TOTAL_FITTED_bkg_events_err'] = pdfFittedErrInRegionList

    # components
    for isam, sample in enumerate(sampleList):
        nSampleInRegionVal = []
        nSampleInRegionError = []
        sampleInAllRegions = RooArgSet()
        for ireg, region in enumerate(regionList):
            sampleInRegion = Util.GetComponent(w, sample, region)
            sampleInRegionVal = 0.
            sampleInRegionError = 0.
            if not sampleInRegion == None:
                sampleInRegion.Print()
                sampleInRegionVal = sampleInRegion.getVal()
                sampleInRegionError = Util.GetPropagatedError(
                    sampleInRegion, resultAfterFit)
                sampleInAllRegions.add(sampleInRegion)
            else:
                print " \n YieldsTable.py: WARNING: sample =", sample, " non-existent (empty) in region =", region, "\n"
            nSampleInRegionVal.append(sampleInRegionVal)
            nSampleInRegionError.append(sampleInRegionError)
        # print " \n\n  XXX-AFTER sample = ", sample
        if showSum:
            sampleSumInAllRegions = RooAddition(
                (sample + "_AllRegions_FITTED"),
                (sample + "_AllRegions_FITTED"), sampleInAllRegions)
            sampleSumInAllRegions.Print()
            nSampleSumVal = sampleSumInAllRegions.getVal()
            nSampleSumError = Util.GetPropagatedError(sampleSumInAllRegions,
                                                      resultAfterFit)
            nSampleInRegionVal.append(nSampleSumVal)
            nSampleInRegionError.append(nSampleSumError)
        tablenumbers['Fitted_events_' + sample] = nSampleInRegionVal
        tablenumbers['Fitted_err_' + sample] = nSampleInRegionError

    ######
    ######
    ######  FROM HERE ON OUT WE CALCULATE THE EXPECTED NUMBER OF EVENTS __BEFORE__ THE FIT
    ######
    ######

    #  FROM HERE ON OUT WE CALCULATE THE EXPECTED NUMBER OF EVENTS BEFORE THE FIT
    w.loadSnapshot('snapshot_paramsVals_RooExpandedFitResult_beforeFit')

    pdfinRegionList = [Util.GetRegionPdf(w, region) for region in regionList]
    varinRegionList = [Util.GetRegionVar(w, region) for region in regionList]
    rrspdfinRegionList = []
    for index, pdf in enumerate(pdfinRegionList):
        prodList = pdf.pdfList()
        foundRRS = 0
        for idx in range(prodList.getSize()):
            if prodList[idx].InheritsFrom("RooRealSumPdf"):
                #      print " \n\n  XXX-BEFORE  prodList[idx] = ", prodList[idx].GetName()
                prodList[idx].Print()
                rrspdfInt = prodList[idx].createIntegral(
                    RooArgSet(varinRegionList[index]))
                rrspdfinRegionList.append(rrspdfInt)
                foundRRS += 1
        if foundRRS > 1 or foundRRS == 0:
            print " \n\n WARNING: ", pdf.GetName(
            ), " has ", foundRRS, " instances of RooRealSumPdf"
            print pdf.GetName(), " component list:", prodList.Print("v")

    nExpInRegionList = [
        pdf.getVal() for index, pdf in enumerate(rrspdfinRegionList)
    ]
    pdfExpErrInRegionList = [
        Util.GetPropagatedError(pdf, resultBeforeFit)
        for pdf in rrspdfinRegionList
    ]

    if showSum:
        pdfInAllRegions = RooArgSet()
        for index, pdf in enumerate(rrspdfinRegionList):
            pdfInAllRegions.add(pdf)
        pdfSumInAllRegions = RooAddition("pdf_AllRegions_BEFORE",
                                         "pdf_AllRegions_BEFORE",
                                         pdfInAllRegions)
        nPdfSumVal = pdfSumInAllRegions.getVal()
        nPdfSumError = Util.GetPropagatedError(pdfSumInAllRegions,
                                               resultAfterFit)
        nExpInRegionList.append(nPdfSumVal)
        pdfExpErrInRegionList.append(nPdfSumError)

    tablenumbers['TOTAL_MC_EXP_BKG_events'] = nExpInRegionList
    tablenumbers['TOTAL_MC_EXP_BKG_err'] = pdfExpErrInRegionList

    for isam, sample in enumerate(sampleList):
        nMCSampleInRegionVal = []
        nMCSampleInRegionError = []
        sampleInAllRegions = RooArgSet()
        for ireg, region in enumerate(regionList):
            MCSampleInRegion = Util.GetComponent(w, sample, region)
            MCSampleInRegionVal = 0.
            MCSampleInRegionError = 0.
            if not MCSampleInRegion == None:
                MCSampleInRegionVal = MCSampleInRegion.getVal()
                MCSampleInRegionError = Util.GetPropagatedError(
                    MCSampleInRegion, resultBeforeFit)
                sampleInAllRegions.add(sampleInRegion)
            else:
                print " \n WARNING: sample=", sample, " non-existent (empty) in region=", region
            nMCSampleInRegionVal.append(MCSampleInRegionVal)
            nMCSampleInRegionError.append(MCSampleInRegionError)
        #print " \n\n  XXX-BEFORE  sample = ", sample
        if showSum:
            sampleSumInAllRegions = RooAddition((sample + "_AllRegions_MC"),
                                                (sample + "_AllRegions_MC"),
                                                sampleInAllRegions)
            nSampleSumVal = sampleSumInAllRegions.getVal()
            nSampleSumError = Util.GetPropagatedError(sampleSumInAllRegions,
                                                      resultBeforeFit)
            nMCSampleInRegionVal.append(nSampleSumVal)
            nMCSampleInRegionError.append(nSampleSumError)
        tablenumbers['MC_exp_events_' + sample] = nMCSampleInRegionVal
        tablenumbers['MC_exp_err_' + sample] = nMCSampleInRegionError

        #  sorted(tablenumbers, key=lambda sample: sample[1])   # sort by age
    map_listofkeys = tablenumbers.keys()
    map_listofkeys.sort()

    for name in map_listofkeys:
        if tablenumbers.has_key(name):
            print name, ": ", tablenumbers[name]

    ###
    return tablenumbers
Exemple #7
0
def latexfitresults(filename, region, sample, resultName, dataname, doAsym):
    """
    Method: set all parameters constant, except for the one you're interested in,
    calculate the systematic/error propagated due to that parameter

    filename: The filename containing afterFit workspace
    resultname: The name of fit result (typically='RooExpandedFitResult_afterFit' or 'RooExpandedFitResult_beforeFit'
    region: The region to be used for systematics breakdown calculation
    sample: The sample to be used insted of total pdf (default='' not defined, hence total pdf used)
    dataname: The name of dataset (default='obsData')
    doAsym: Calculates asymmetric errors taken from MINOS (default=True)
    """

    #pick up workspace from file
    workspacename = 'w'
    w = Util.GetWorkspaceFromFile(filename,workspacename)
    if w is None:
        print "ERROR : Cannot open workspace:", workspacename
        sys.exit(1)

    #pickup RooExpandedFitResult from workspace with name resultName (either before or after fit)
    result = w.obj(resultName)
    if result is None:
        print "ERROR : Cannot open fit result", resultName
        sys.exit(1)

    # load workspace snapshot related to resultName (=set all parameters to values after fit)
    snapshot =  'snapshot_paramsVals_' + resultName
    w.loadSnapshot(snapshot)

    # pick up dataset from workspace
    data_set = w.data(dataname)
    if data_set is None:
        print "ERROR : Cannot open dataset: ", "data_set"
        sys.exit(1)

    #pick up channel category (RooCategory) from workspace
    region_cat = w.obj("channelCat")
    data_set.table(region_cat).Print("v");

    # find full (long) name list of region (i.e. short=SR3J, long=SR3J_meffInc30_JVF25pt50)
    region_full_name = Util.GetFullRegionName(region_cat, region);

    # set a boolean whether we're looking at a sample or the full (multi-sample) pdf/model
    chosen_sample = bool(sample)

    # define regSys set, for all names/numbers to be saved in
    reg_sys = {}

    # define channelCat call for this region and reduce the dataset to this category/region
    region_cat_str = 'channelCat==channelCat::' + region_full_name.Data()
    data_region = data_set.reduce(region_cat_str)

    # retrieve and save number of observed (=data) events in region
    nobs_region = 0.
    if data_region:
        nobs_region = data_region.sumEntries()
    else:
        print " ERROR : dataset-category dataRegion not found"

    # if looking at a sample, there is no equivalent N_obs (only for the full model)
    if chosen_sample:
        reg_sys['sqrtnobsa'] = 0.
    else:
        reg_sys['sqrtnobsa'] = ROOT.TMath.Sqrt(nobs_region)

    # get the pdf for the total model or just for the sample in region
    if chosen_sample:
        pdf_in_region = getPdfInRegions(w, sample, region)
    else:
        raw_pdf_in_region = Util.GetRegionPdf(w, region)
        var_in_region =  Util.GetRegionVar(w, region)
        prod_list = raw_pdf_in_region.pdfList()

        foundRRS = 0
        for idx in xrange(prod_list.getSize()):
            if prod_list[idx].InheritsFrom("RooRealSumPdf"):
                rrspdf_int =  prod_list[idx].createIntegral(ROOT.RooArgSet(var_in_region));
                pdf_in_region = rrspdf_int
                foundRRS += 1

        if foundRRS > 1 or foundRRS == 0:
            print " \n\n WARNING: ", pdf.GetName(), " has ", foundRRS, " instances of RooRealSumPdf"
            print pdf.GetName(), " component list:", prodList.Print("v")

        if not pdf_in_region:
            if chosenSample:
                print " \n Warning, could not find pdf in region = ",region, " for sample = ",sample
            else:
                print " \n Warning, could not find pdf in region = ",region

    # calculate fitted pdf number of events and full error
    n_fitted_in_region = pdf_in_region.getVal()
    reg_sys['sqrtnfitted'] = ROOT.TMath.Sqrt(n_fitted_in_region)
    reg_sys['nfitted'] = n_fitted_in_region

    pdf_fitted_err_in_region = Util.GetPropagatedError(pdf_in_region, result, doAsym)
    reg_sys['totsyserr'] = pdf_fitted_err_in_region


    # calculate error per (floating) parameter in fitresult
    # get a list of floating parameters to loop over
    fpf = result.floatParsFinal()

    # set all floating parameters constant
    for idx in xrange(fpf.getSize()):
        parname = fpf[idx].GetName()
        par = w.var(parname)
        par.setConstant(True)

    for idx in xrange(fpf.getSize()):
        parname = fpf[idx].GetName()
        par = w.var(parname)
        par.setConstant(False)
        reg_sys['syserr_'+parname] =  Util.GetPropagatedError(pdf_in_region, result, doAsym)
        par.setConstant(True)

    return reg_sys
def latexfitresults(filename, regionList, dataname='obsData'):

    print "hallo"
    ####

    w = Util.GetWorkspaceFromFile(filename, 'combined')

    if w == None:
        print "ERROR : Cannot open workspace : ", workspacename
        sys.exit(1)

    w.Print()

    resultname = 'RooExpandedFitResult_afterFit'
    result = w.obj(resultname)

    if result == None:
        print "ERROR : Cannot open fit result : ", resultname
        sys.exit(1)

    snapshot = 'snapshot_paramsVals_' + resultname
    w.loadSnapshot(snapshot)

    if not w.loadSnapshot(snapshot):
        print "ERROR : Cannot load snapshot : ", snapshot
        sys.exit(1)

    data_set = w.data(dataname)
    if data_set == None:
        print "ERROR : Cannot open dataset : ", "data_set" + suffix
        sys.exit(1)

    regionCat = w.obj("channelCat")
    data_set.table(regionCat).Print("v")

    regionFullNameList = [
        Util.GetFullRegionName(regionCat, region) for region in regionList
    ]
    print " \n Requested regions = ", regionFullNameList, "\n\n"

    ###

    tablenumbers = {}
    tablenumbers['names'] = regionList

    regionCatList = [
        'channelCat==channelCat::' + region.Data()
        for region in regionFullNameList
    ]

    regionDatasetList = [
        data_set.reduce(regioncat) for regioncat in regionCatList
    ]
    for index, data in enumerate(regionDatasetList):
        data.SetName("data_" + regionList[index])
        data.SetTitle("data_" + regionList[index])

    nobs_regionList = [data.sumEntries() for data in regionDatasetList]
    tablenumbers['nobs'] = nobs_regionList

    ####

    bkginRegionList = [
        Util.GetComponent(w, "Top,WZ,QCD,BG", region) for region in regionList
    ]
    nbkginRegionList = [
        bkginRegion.getVal() for bkginRegion in bkginRegionList
    ]
    [region.Print() for region in bkginRegionList]
    print "\n N bkgs in regions = ", nbkginRegionList

    nbkgerrinRegionList = [
        Util.GetPropagatedError(bkginRegion, result)
        for bkginRegion in bkginRegionList
    ]
    print "\n error N bkgs in regions = ", nbkgerrinRegionList

    WZinRegionList = [
        Util.GetComponent(w, "WZ", region) for region in regionList
    ]
    TopinRegionList = [
        Util.GetComponent(w, "Top", region) for region in regionList
    ]

    nWZinRegionList = [WZinRegion.getVal() for WZinRegion in WZinRegionList]
    nTopinRegionList = [
        TopinRegion.getVal() for TopinRegion in TopinRegionList
    ]

    print "\n N WZ in regions = ", nWZinRegionList
    print "\n N Top in regions = ", nTopinRegionList

    nWZerrinRegionList = [
        Util.GetPropagatedError(WZinRegion, result)
        for WZinRegion in WZinRegionList
    ]
    nToperrinRegionList = [
        Util.GetPropagatedError(TopinRegion, result)
        for TopinRegion in TopinRegionList
    ]

    print "\n error N WZ in regions = ", nWZerrinRegionList
    print "\n error N Top in regions = ", nToperrinRegionList

    ######
    # Example how to add multiple backgrounds in multiple(!) regions
    TopWZinRegionList = [
        RooAddition(("TopWZin" + regionList[index]),
                    ("TopWZin" + regionList[index]),
                    RooArgSet(TopinRegionList[index], WZinRegionList[index]))
        for index, region in enumerate(regionList)
    ]
    nTopWZinRegionList = [
        TopWZinRegion.getVal() for TopWZinRegion in TopWZinRegionList
    ]
    nTopWZerrinRegionList = [
        Util.GetPropagatedError(TopWZinRegion, result)
        for TopWZinRegion in TopWZinRegionList
    ]

    ######
    ######
    ######  FROM HERE ON OUT WE CALCULATE THE EXPECTED NUMBER OF EVENTS __BEFORE__ THE FIT
    ######
    ######

    #  FROM HERE ON OUT WE CALCULATE THE EXPECTED NUMBER OF EVENTS BEFORE THE FIT
    w.loadSnapshot('snapshot_paramsVals_RooExpandedFitResult_beforeFit')

    pdfinRegionList = [Util.GetRegionPdf(w, region) for region in regionList]

    varinRegionList = [Util.GetRegionVar(w, region) for region in regionList]

    nexpinRegionList = [
        pdf.expectedEvents(RooArgSet(varinRegionList[index]))
        for index, pdf in enumerate(pdfinRegionList)
    ]
    print "\n N expected in regions = ", nexpinRegionList

    fracWZinRegionList = [
        Util.GetComponentFracInRegion(w, "WZ", region) for region in regionList
    ]
    fracTopinRegionList = [
        Util.GetComponentFracInRegion(w, "Top", region)
        for region in regionList
    ]

    mcWZinRegionList = [
        fracWZinRegionList[index] * nexpinRegionList[index]
        for index, region in enumerate(regionList)
    ]
    mcTopinRegionList = [
        fracTopinRegionList[index] * nexpinRegionList[index]
        for index, region in enumerate(regionList)
    ]

    #  mcSMinRegionList = [ mcWZinRegionList[index] + mcTopinRegionList[index] +  mcQCDinRegionList[index] + mcBGinRegionList[index]  for index, region in enumerate(regionList)]
    mcSMinRegionList = [
        mcWZinRegionList[index] + mcTopinRegionList[index]
        for index, region in enumerate(regionList)
    ]
    print "\n N expected WZ in regions = ", mcWZinRegionList
    print "\n N expected Top in regions = ", mcTopinRegionList

    tablenumbers['MC_exp_top_WZ_events'] = [
        mcWZinRegionList[index] + mcTopinRegionList[index]
        for index, region in enumerate(regionList)
    ]
    tablenumbers['MC_exp_top_events'] = [
        mcTopinRegion for mcTopinRegion in mcTopinRegionList
    ]
    tablenumbers['MC_exp_WZ_events'] = [
        mcWZinRegion for mcWZinRegion in mcWZinRegionList
    ]
    tablenumbers['MC_exp_SM_events'] = [
        mcSMinRegion for mcSMinRegion in mcSMinRegionList
    ]

    tablenumbers['Fitted_bkg_events'] = nbkginRegionList
    tablenumbers['Fitted_bkg_events_err'] = nbkgerrinRegionList

    tablenumbers['Fitted_top_events'] = nTopinRegionList
    tablenumbers['Fitted_top_events_err'] = nToperrinRegionList

    tablenumbers['Fitted_WZ_events'] = nWZinRegionList
    tablenumbers['Fitted_WZ_events_err'] = nWZerrinRegionList

    tablenumbers['Fitted_top_WZ_events'] = nTopWZinRegionList
    tablenumbers['Fitted_top_WZ_events_err'] = nTopWZerrinRegionList

    ###
    return tablenumbers
Exemple #9
0
def latexfitresults(filename, region_list, sample_list):

    f = ROOT.TFile.Open(filename)
    w = f.Get('w')

    doAsym = True

    if w is None:
        print "ERROR : Cannot open workspace : w"
        sys.exit(1)

    resultAfterFit = w.obj('RooExpandedFitResult_afterFit')
    if resultAfterFit is None:
        print(
            "ERROR : Cannot open fit result after fit RooExpandedFitResult_afterFit"
        )
        sys.exit(1)

    resultBeforeFit = w.obj('RooExpandedFitResult_beforeFit')
    if resultBeforeFit is None:
        print(
            "ERROR : Cannot open fit result before fit RooExpandedFitResult_beforeFit"
        )
        sys.exit(1)

    # pick up dataset from workspace
    data_set = w.data('obsData')

    # pick up channel category (RooCategory) from workspace
    regionCat = w.obj("channelCat")

    # find full (long) name list of regions (i.e. short=SR3J, long=SR3J_meffInc30_JVF25pt50)
    regionFullNameList = [
        Util.GetFullRegionName(regionCat, region) for region in region_list
    ]

    # load afterFit workspace snapshot (=set all parameters to values after fit)
    snapshot = 'snapshot_paramsVals_RooExpandedFitResult_afterFit'
    w.loadSnapshot(snapshot)

    if not w.loadSnapshot(snapshot):
        print "ERROR : Cannot load snapshot : ", snapshot
        sys.exit(1)

    # define set, for all names/yields to be saved in
    tablenumbers = dict()

    tablenumbers['names'] = region_list

    # make a list of channelCat calls for every region
    regionCatList = [
        'channelCat==channelCat::' + region.Data()
        for region in regionFullNameList
    ]

    # retrieve number of observed (=data) events per region
    regionDatasetList = [
        data_set.reduce(regioncat) for regioncat in regionCatList
    ]
    for index, data in enumerate(regionDatasetList):
        data.SetName("data_" + region_list[index])
        data.SetTitle("data_" + region_list[index])

    nobs_regionList = [data.sumEntries() for data in regionDatasetList]

    tablenumbers['nobs'] = nobs_regionList

    # FROM HERE ON OUT WE CALCULATE THE FITTED NUMBER OF EVENTS __AFTER__ THE FIT

    #get a list of pdf's and variables per region
    pdfinRegionList = [Util.GetRegionPdf(w, region) for region in region_list]
    varinRegionList = [Util.GetRegionVar(w, region) for region in region_list]

    varNbinsInRegionList = []
    varBinLowInRegionList = []
    varBinHighInRegionList = []
    rangeNameBinsInRegionList = []

    #  get a list of RooRealSumPdf per region (RooRealSumPdf is the top-pdf per region containing all samples)
    rrspdfinRegionList = []
    for index, pdf in enumerate(pdfinRegionList):
        if not pdf:
            print "WARNING: pdf is NULL for index {0}".format(index)
            continue

        prodList = pdf.pdfList()
        foundRRS = 0

        for idx in range(prodList.getSize()):
            if prodList[idx].InheritsFrom("RooRealSumPdf"):
                rrspdfInt = prodList[idx].createIntegral(
                    ROOT.RooArgSet(varinRegionList[index]))
                rrspdfinRegionList.append(rrspdfInt)
                foundRRS += 1

        if foundRRS > 1 or foundRRS == 0:
            print " \n\n WARNING: ", pdf.GetName(
            ), " has ", foundRRS, " instances of RooRealSumPdf"
            print pdf.GetName(), " component list:", prodList.Print("v")

    # calculate total pdf number of fitted events and error
    nFittedInRegionList = [
        pdf.getVal() for index, pdf in enumerate(rrspdfinRegionList)
    ]
    pdfFittedErrInRegionList = [
        Util.GetPropagatedError(pdf, resultAfterFit, doAsym)
        for pdf in rrspdfinRegionList
    ]

    tablenumbers['TOTAL_FITTED_bkg_events'] = nFittedInRegionList
    tablenumbers['TOTAL_FITTED_bkg_events_err'] = pdfFittedErrInRegionList

    # calculate the fitted number of events and propagated error for each requested sample, by splitting off each sample pdf
    for isam, sample in enumerate(sample_list):

        sampleName = getName(sample)

        nSampleInRegionVal = []
        nSampleInRegionError = []
        sampleInAllRegions = ROOT.RooArgSet()

        for ireg, region in enumerate(region_list):
            sampleInRegion = getPdfInRegions(w, sample, region)
            sampleInRegionVal = 0.
            sampleInRegionError = 0.

            try:
                sampleInRegionVal = sampleInRegion.getVal()
                sampleInRegionError = Util.GetPropagatedError(
                    sampleInRegion, resultAfterFit, doAsym)
                sampleInAllRegions.add(sampleInRegion)
            except:
                print " \n YieldsTable.py: WARNING: sample =", sampleName, " non-existent (empty) in region =", region, "\n"

            nSampleInRegionVal.append(sampleInRegionVal)
            nSampleInRegionError.append(sampleInRegionError)

        tablenumbers['Fitted_events_' + sampleName] = nSampleInRegionVal
        tablenumbers['Fitted_err_' + sampleName] = nSampleInRegionError

    # FROM HERE ON OUT WE CALCULATE THE EXPECTED NUMBER OF EVENTS __BEFORRE__ THE FIT

    #load beforeFit workspace snapshot (=set all parameters to values before fit)
    w.loadSnapshot('snapshot_paramsVals_RooExpandedFitResult_beforeFit')

    # check if any of the initial scaling factors is != 1
    _result = w.obj('RooExpandedFitResult_beforeFit')
    _muFacs = _result.floatParsFinal()

    for i in xrange(len(_muFacs)):

        if "mu_" in _muFacs[i].GetName() and _muFacs[i].getVal() != 1.0:
            print " \n WARNING: scaling factor %s != 1.0 (%g) expected MC yield WILL BE WRONG!" % (
                _muFacs[i].GetName(), _muFacs[i].getVal())

    # get a list of pdf's and variables per region
    pdfinRegionList = [Util.GetRegionPdf(w, region) for region in region_list]
    varinRegionList = [Util.GetRegionVar(w, region) for region in region_list]

    # get a list of RooRealSumPdf per region (RooRealSumPdf is the top-pdf per region containing all samples)
    rrspdfinRegionList = []
    for index, pdf in enumerate(pdfinRegionList):
        if not pdf:
            print "WARNING: pdf is NULL for index {0}".format(index)
            continue
        prodList = pdf.pdfList()
        foundRRS = 0
        for idx in xrange(prodList.getSize()):
            if prodList[idx].InheritsFrom("RooRealSumPdf"):
                rrspdfInt = prodList[idx].createIntegral(
                    ROOT.RooArgSet(varinRegionList[index]))
                rrspdfinRegionList.append(rrspdfInt)
                foundRRS += 1

        if foundRRS > 1 or foundRRS == 0:
            print " \n\n WARNING: ", pdf.GetName(
            ), " has ", foundRRS, " instances of RooRealSumPdf"
            print pdf.GetName(), " component list:", prodList.Print("v")

    # calculate total pdf number of expected events and error
    nExpInRegionList = [
        pdf.getVal() for index, pdf in enumerate(rrspdfinRegionList)
    ]
    pdfExpErrInRegionList = [
        Util.GetPropagatedError(pdf, resultBeforeFit, doAsym)
        for pdf in rrspdfinRegionList
    ]

    tablenumbers['TOTAL_MC_EXP_BKG_events'] = nExpInRegionList
    tablenumbers['TOTAL_MC_EXP_BKG_err'] = pdfExpErrInRegionList

    # calculate the fitted number of events and propagated error for each requested sample, by splitting off each sample pdf
    for isam, sample in enumerate(sample_list):

        sampleName = getName(sample)

        nMCSampleInRegionVal = []
        nMCSampleInRegionError = []
        MCSampleInAllRegions = ROOT.RooArgSet()

        for ireg, region in enumerate(region_list):
            MCSampleInRegion = getPdfInRegions(w, sample, region)
            MCSampleInRegionVal = 0.
            MCSampleInRegionError = 0.

            try:
                MCSampleInRegionVal = MCSampleInRegion.getVal()
                MCSampleInRegionError = Util.GetPropagatedError(
                    MCSampleInRegion, resultBeforeFit, doAsym)
                MCSampleInAllRegions.add(MCSampleInRegion)
            except:
                print " \n WARNING: sample=", sampleName, " non-existent (empty) in region=", region

            nMCSampleInRegionVal.append(MCSampleInRegionVal)
            nMCSampleInRegionError.append(MCSampleInRegionError)

        tablenumbers['MC_exp_events_' + sampleName] = nMCSampleInRegionVal
        tablenumbers['MC_exp_err_' + sampleName] = nMCSampleInRegionError

    # sort the tablenumbers set
    map_listofkeys = tablenumbers.keys()
    map_listofkeys.sort()

    f.Close()

    return tablenumbers