def runLikelihood(subdir, tpl_file): '''This runction runs the likelihood code on a set of pixels in a subdirectory. It takes as input the subdirectory to work on and a template counts map. It reads it's configuration from a pickle file (par.pck) that should be located in the subdirectory and the pixel locations from another pickle file (pixel.pck). It then creats an overall likelihood object, does a quick global fit and then loops over the pixels. At each pixel, it creats a test source, fits that source, calculates the TS of the source and writes the results to an output file in the subdirectory called ts_results.dat.''' parfile = open("par.pck", "r") pars = pickle.load(parfile) pixelfile = open("pixel.pck", "r") pixels = pickle.load(pixelfile) pixel_coords = PixelCoords(tpl_file) obs = UnbinnedObs(resolve_fits_files(pars['evfile']), resolve_fits_files(pars['scfile']), expMap='../'+pars['expmap'], expCube='../'+pars['expcube'], irfs=pars['irfs']) like = UnbinnedAnalysis(obs, '../'+pars['srcmdl'], pars['optimizer']) like.setFitTolType(pars['toltype']) like.optimize(0) loglike0 = like() test_src = getPointSource(like) target_name = 'testSource' test_src.setName(target_name) outfile = 'ts_results.dat' finished_pixels = [] if os.path.isfile(outfile): input = open(outfile, 'r') for line in input: tokens = line.strip().split() ij = int(tokens[0]), int(tokens[1]) finished_pixels.append(ij) input.close() output = open(outfile, 'a') for indx, i, j in pixels: if (i, j) in finished_pixels: continue ra, dec = pixel_coords(i, j) test_src.setDir(ra, dec, True, False) like.addSource(test_src) like.optimize(0) ts = -2*(like() - loglike0) output.write("%3i %3i %.3f %.3f %.5f\n" % (i, j, ra, dec, ts)) output.flush() like.deleteSource(target_name) output.close()
def runLikelihood(subdir, tpl_file): '''This runction runs the likelihood code on a set of pixels in a subdirectory. It takes as input the subdirectory to work on and a template counts map. It reads it's configuration from a pickle file (par.pck) that should be located in the subdirectory and the pixel locations from another pickle file (pixel.pck). It then creats an overall likelihood object, does a quick global fit and then loops over the pixels. At each pixel, it creats a test source, fits that source, calculates the TS of the source and writes the results to an output file in the subdirectory called ts_results.dat.''' parfile = open("par.pck", "r") pars = pickle.load(parfile) pixelfile = open("pixel.pck", "r") pixels = pickle.load(pixelfile) pixel_coords = PixelCoords(tpl_file) obs = UnbinnedObs(resolve_fits_files(pars['evfile']), resolve_fits_files(pars['scfile']), expMap='../' + pars['expmap'], expCube='../' + pars['expcube'], irfs=pars['irfs']) like = UnbinnedAnalysis(obs, '../' + pars['srcmdl'], pars['optimizer']) like.setFitTolType(pars['toltype']) like.optimize(0) loglike0 = like() test_src = getPointSource(like) target_name = 'testSource' test_src.setName(target_name) outfile = 'ts_results.dat' finished_pixels = [] if os.path.isfile(outfile): input = open(outfile, 'r') for line in input: tokens = line.strip().split() ij = int(tokens[0]), int(tokens[1]) finished_pixels.append(ij) input.close() output = open(outfile, 'a') for indx, i, j in pixels: if (i, j) in finished_pixels: continue ra, dec = pixel_coords(i, j) test_src.setDir(ra, dec, True, False) like.addSource(test_src) like.optimize(0) ts = -2 * (like() - loglike0) output.write("%3i %3i %.3f %.3f %.5f\n" % (i, j, ra, dec, ts)) output.flush() like.deleteSource(target_name) output.close()
# N.B A source with a TS value less than 0 should never happen unless the minimization failed. Remove that source and try fitting again and check the return code again. converganceCheckOne="Zero?: %d" %(likeobj.getRetCode()) converganceCheckTwo="NA" print converganceCheckOne # This value should be zero, refit using the method below if not if not likeobj.getRetCode() ==0 or eMin<100000: # If energy is too high, just don't refit because it causes problems print colors.OKBLUE+"Minimization Failed. Removing sources with TS < 9.0 and refitting!" # Protip, you can really simplify the model by removing sources with TS levels below 9 (about 3 sigma) for source,TS in sourceDetails.iteritems(): print source, TS if (TS < TSThreshold): print "Deleting..." like.deleteSource(source) # Refit like.fit(verbosity=0,covar=True,optObject=likeobj) converganceCheckTwo="Zero?: %d" %(likeobj.getRetCode()) print converganceCheckTwo sourceDetails = {} for source in like.sourceNames(): sourceDetails[source] = like.Ts(source) print colors.ENDC printDictionaryToFile(sourceDetails,'TSValues.txt') ########################################
def main(NAME, RA, DEC, TSTART, TSTOP, EMIN, EMAX, SC, ROIu, xml): ROIue = float(ROIu) + 10 os.system('ls -1 *PH*.fits > %s_events.list' % (NAME)) my_apps.filter['evclass'] = 128 my_apps.filter['evtype'] = 3 my_apps.filter['ra'] = RA my_apps.filter['dec'] = DEC my_apps.filter['rad'] = ROIu my_apps.filter['emin'] = EMIN my_apps.filter['emax'] = EMAX my_apps.filter['zmax'] = 90 my_apps.filter['tmin'] = TSTART my_apps.filter['tmax'] = TSTOP my_apps.filter['infile'] = '@%s_events.list' % (NAME) my_apps.filter['outfile'] = '%s_filtered.fits' % (NAME) my_apps.filter.run() # maketime my_apps.maketime['scfile'] = SC my_apps.maketime['filter'] = '(DATA_QUAL>0)&&(LAT_CONFIG==1)' my_apps.maketime['roicut'] = 'no' my_apps.maketime['evfile'] = '%s_filtered.fits' % (NAME) my_apps.maketime['outfile'] = '%s_filtered_gti.fits' % (NAME) my_apps.maketime.run() # my_apps.counts_map['evfile'] = '%s_filtered_gti.fits' % (NAME) my_apps.counts_map['scfile'] = SC my_apps.counts_map['outfile'] = '%s_CountMap.fits' % (NAME) # my_apps.counts_map.run() # my_apps.expCube['evfile'] = '%s_filtered_gti.fits' % (NAME) my_apps.expCube['scfile'] = SC my_apps.expCube['zmax'] = 90 my_apps.expCube['outfile'] = 'expCube.fits' my_apps.expCube['dcostheta'] = 0.025 my_apps.expCube['binsz'] = 1 my_apps.expCube.run() my_apps.expMap['evfile'] = '%s_filtered_gti.fits' % (NAME) my_apps.expMap['scfile'] = SC my_apps.expMap['expcube'] = 'expCube.fits' my_apps.expMap['outfile'] = 'expMap.fits' my_apps.expMap['irfs'] = 'CALDB' my_apps.expMap['srcrad'] = ROIue my_apps.expMap['nlong'] = 120 my_apps.expMap['nlat'] = 120 my_apps.expMap['nenergies'] = 20 my_apps.expMap.run() # sara xml model roiname = '%s_filtered_gti.fits' % NAME if float(xml) == 0: xml_creator_P8_v1.main(NAME, float(RA), float(DEC), float(EMIN), float(EMAX), 15) xmlmodelname = '%s_model.xml' % NAME my_apps.diffResps['evfile'] = '%s_filtered_gti.fits' % (NAME) my_apps.diffResps['scfile'] = SC my_apps.diffResps['srcmdl'] = xmlmodelname my_apps.diffResps['irfs'] = 'CALDB' my_apps.diffResps.run() xmlfitname = '%s_fit1.xml' % NAME obs = UnbinnedObs(roiname, SC, expMap='expMap.fits', expCube='expCube.fits', irfs='CALDB') # like1 = UnbinnedAnalysis(obs,xmlmodelname,optimizer='MINUIT') like1 = UnbinnedAnalysis(obs, xmlmodelname, optimizer='NewMinuit') likeobj = pyLike.NewMinuit(like1.logLike) like1.fit(verbosity=0, optObject=likeobj) print likeobj.getRetCode() sourceDetails = {} for source in like1.sourceNames(): sourceDetails[source] = like1.Ts(source) for source, TS in sourceDetails.iteritems(): if (TS < 2): print "Deleting...", source, " TS = ", TS like1.deleteSource(source) like1.fit(verbosity=0, optObject=likeobj) print "0 is converged", likeobj.getRetCode() like1.logLike.writeXml(xmlfitname) numl = search(NAME, xmlfitname) numlg = str(numl + 3) os.system("sed '" + numlg + "," + numlg + " s/free=\"1\"/free=\"0\"/' " + xmlfitname + " > xml_sed.xml ") inputs = likeInput(like1, NAME, model="xml_sed.xml", nbins=6, phCorr=1.0) inputs.plotBins() inputs.fullFit(CoVar=True) sed = likeSED(inputs) sed.getECent() sed.fitBands() sed.Plot() result = like1.model[NAME] TS = like1.Ts(NAME) flux = like1.flux(NAME, emin=100) gamma = like1.model[NAME].funcs['Spectrum'].getParam('Index').value() cov_gg = like1.model[NAME].funcs['Spectrum'].getParam('Index').error() # cov_II = like1.model[NAME].funcs['Spectrum'].getParam('Integral').error() flux_err = like1.fluxError(NAME, emin=100) like1.plot() fitsedname = '%s_6bins_likeSEDout.fits' % NAME sedtool(fitsedname) print NAME, " TS=", TS print result if float(xml) == 1: xmlmodelname = '%s_model.xml' % NAME xmlfitname = '%s_fit1.xml' % NAME obs = UnbinnedObs(roiname, SC, expMap='expMap.fits', expCube='expCube.fits', irfs='CALDB') # like1 = UnbinnedAnalysis(obs,xmlmodelname,optimizer='MINUIT') like1 = UnbinnedAnalysis(obs, xmlmodelname, optimizer='NewMinuit') likeobj = pyLike.NewMinuit(like1.logLike) like1.fit(verbosity=0, optObject=likeobj) print likeobj.getRetCode() sourceDetails = {} for source in like1.sourceNames(): sourceDetails[source] = like1.Ts(source) for source, TS in sourceDetails.iteritems(): if (TS < 2): print "Deleting...", source, " TS = ", TS like1.deleteSource(source) like1.fit(verbosity=0, optObject=likeobj) print "0 is converged", likeobj.getRetCode() like1.logLike.writeXml(xmlfitname) numl = search(NAME, xmlfitname) numlg = str(numl + 3) os.system("sed '" + numlg + "," + numlg + " s/free=\"1\"/free=\"0\"/' " + xmlfitname + " > xml_sed.xml ") inputs = likeInput(like1, NAME, model="xml_sed.xml", nbins=6, phCorr=1.0) inputs.plotBins() inputs.fullFit(CoVar=True) sed = likeSED(inputs) sed.getECent() sed.fitBands() sed.Plot() result = like1.model[NAME] TS = like1.Ts(NAME) flux = like1.flux(NAME, emin=100) gamma = like1.model[NAME].funcs['Spectrum'].getParam('Index').value() cov_gg = like1.model[NAME].funcs['Spectrum'].getParam('Index').error() # cov_II = like1.model[NAME].funcs['Spectrum'].getParam('Integral').error() flux_err = like1.fluxError(NAME, emin=100) like1.plot() fitsedname = '%s_6bins_likeSEDout.fits' % NAME sedtool(fitsedname) print NAME, " TS=", TS print result
class quickLike: """ This is the base class. A usual likelihood analysis will consists of running the following functions (assuming you have a configuration file): * qL = quickLike('MySource', True) * qL.makeObs() * qL.initDRM() * qL.fitDRM() * qL.initMIN() * qL.fitMIN() This will set up all of the objects needed for the analysis and do an initial fit with one of the DRM optimizers. It'll save these results and use them for the second fit with one of the Minuit optimizers. If you do not have a configuration file, you'll need to input all of the options for this module when you create the quickLike object (see the various options below). You can create a configuration file by executing writeConfig(). * qL.writeConfig() This module will catch any failures from the optimizers and will report them to the user. There are a few functions that are useful to use in this case:""" def __init__(self, base = 'MySource', configFile = False, likelihoodConfig = {"model" : "MySource_model.xml", "sourcename" : "Source Name", "drmtol" : 0.1, "mintol" : 1e-4}, commonConfig = {"base" : 'MySource', "eventclass" : 2, "binned" : False, "irfs" : "P7SOURCE_V6", "verbosity" : 0}): commonConfig['base'] = base self.logger = initLogger(base, 'quickLike') if(configFile): try: commonConfigRead,analysisConfigRead,likelihoodConfigRead,plotConfigRead = readConfig(self.logger,base) except(FileNotFound): self.logger.critical("One or more needed files do not exist") return try: commonConfig = checkConfig(self.logger,commonConfig,commonConfigRead) except(KeyError): return try: likelihoodConfig = checkConfig(self.logger,likelihoodConfig,likelihoodConfigRead) except(KeyError): return self.commonConf = commonConfig self.likelihoodConf = likelihoodConfig self.ret = re.compile('\n') self.fitbit = False self.Print() def writeConfig(self): """Writes all of the initialization variables to the config file called <basename>.cfg""" writeConfig(quickLogger=self.logger, commonDictionary=self.commonConf, likelihoodDictionary=self.likelihoodConf) def Print(self): """Prints out information about the various objects to the terminal and to the log file.""" logString = "Created quickLike object: " for variable, value in self.commonConf.iteritems(): logString += variable+"="+str(value)+"," for variable, value in self.likelihoodConf.iteritems(): logString += variable+"="+str(value)+"," self.logger.info(logString) def makeObs(self): """Creates either a binned or unbinned observation object for use in the likelihood analysis. This function checks for all of the needed files first. If you do not have a needed file, see the quickAnalysis module for creation. This function should be run before any of the init or fit functions.""" if(self.commonConf['binned']): try: checkForFiles(self.logger,[self.commonConf['base']+'_srcMaps.fits', self.commonConf['base']+'_ltcube.fits', self.commonConf['base']+'_BinnedExpMap.fits']) self.obs = BinnedObs(srcMaps=self.commonConf['base']+'_srcMaps.fits', expCube=self.commonConf['base']+'_ltcube.fits', binnedExpMap=self.commonConf['base']+'_BinnedExpMap.fits', irfs=self.commonConf['irfs']) except(FileNotFound): self.logger.critical("One or more needed files do not exist") return else: try: checkForFiles(self.logger,[self.commonConf['base']+'_filtered_gti.fits', self.commonConf['base']+'_SC.fits', self.commonConf['base']+'_expMap.fits', self.commonConf['base']+'_ltcube.fits']) self.obs = UnbinnedObs(self.commonConf['base']+'_filtered_gti.fits', self.commonConf['base']+'_SC.fits', expMap=self.commonConf['base']+'_expMap.fits', expCube=self.commonConf['base']+'_ltcube.fits', irfs=self.commonConf['irfs']) except(FileNotFound): self.logger.critical("One or more needed files do not exist") return self.logger.info(self.ret.subn(', ',str(self.obs))[0]) def initDRM(self): """Initializes the DRM optimizer (either binned or unbinned). This is usually the second function that you run when using this module. You need to run makeObs before you run this function. If it hasn't been run, this function will exit.""" try: self.obs except AttributeError: self.logger.critical("Obs object does not exist. Create it first with the makeObs function") return try: checkForFiles(self.logger,[self.likelihoodConf['model']]) if(self.commonConf['binned']): self.DRM = BinnedAnalysis(self.obs,self.likelihoodConf['model'],optimizer="DRMNGB") else: self.DRM = UnbinnedAnalysis(self.obs,self.likelihoodConf['model'],optimizer="DRMNGB") self.DRM.tol = float(self.likelihoodConf['drmtol']) self.logger.info(self.ret.subn(', ',str(self.DRM))[0]) except(FileNotFound): self.logger.critical("One or more needed files do not exist") return def initAltFit(self,opt="MINUIT"): """Initiallizes a minuit optimizer to use as a backup to the DRM optimizer. This function is used internally in the fitDRM function so you probably will never use it. You need to run makeObs before you run this function. If it hasn't been run, this function will exit.""" try: self.obs except AttributeError: self.logger.critical("Obs object does not exist. Create it first with the makeObs function") return try: checkForFiles(self.logger,[self.likelihoodConf['model']]) if(self.commonConf['binned']): self.ALTFIT = BinnedAnalysis(self.obs,self.likelihoodConf['model'],optimizer=opt) else: self.ALTFIT = UnbinnedAnalysis(self.obs,self.likelihoodConf['model'],optimizer=opt) self.ALTFIT.tol = float(self.likelihoodConf['drmtol']) self.ALTFITobj = pyLike.Minuit(self.ALTFIT.logLike) self.logger.info(self.ret.subn(', ',str(self.ALTFIT))[0]) except(FileNotFound): self.logger.critical("One or more needed files do not exist") return def initMIN(self, useBadFit=False): """Initiallizes a New Minuit optimizer to use as a backup to the DRM optimizer. This is usually run after you have initially run fitDRM and created a <basename>_likeDRM.xml model file which is used a seed for the New Minuit optimizer. You can skip the DRM process if you like but you need to have the proper model file (<basename>_likeDRM.xml) present in the working directory. You need to run makeObs before you run this function. If it hasn't been run, this function will exit. If you want to use the non convergant fit from fitDRM, set useBadFit to True.""" try: self.obs except AttributeError: self.logger.critical("Obs object does not exist. Create it first with the makeObs function.") return if(useBadFit): model = self.commonConf['base']+'_badDRMFit.xml' else: model = self.commonConf['base']+'_likeDRM.xml' try: checkForFiles(self.logger,[model]) if(self.commonConf['binned']): self.MIN = BinnedAnalysis(self.obs,model,optimizer='NewMinuit') else: self.MIN = UnbinnedAnalysis(self.obs,model,optimizer='NewMinuit') self.MIN.tol = float(self.likelihoodConf['mintol']) self.MINobj = pyLike.NewMinuit(self.MIN.logLike) self.logger.info(self.ret.subn(', ',str(self.MIN))[0]) except(FileNotFound): self.logger.critical("One or more needed files do not exist") return def fitDRM(self): """Performs a DRM inital fit on your data using the <basename>_model.xml model file. It tries an intial fit and if that fails, tries a tighter tolerance. If that fails, it tries a looser tolerance. If that fails, it tries to do this initial fit with the MINUIT optimizer. If that fails, this function bails. If the fit converges, it saves the results to <basename>_likeDRM.xml which will be used in the NewMinuit fit. If no fit is found, it will save the results to <basename>_badDRMFit.xml. You can use this in the NewMinuit fit if you use the useBadFit option in initMIN. You need to have run initDRM before you run this function.""" try: self.DRM except AttributeError: self.logger.critical("DRM object does not exist. Create it first with the initDRM function.") return altfit=False try: self.DRM.fit(verbosity=int(self.commonConf['verbosity'])) except: self.logger.error("Initial DRM Fit Failed") try: self.logger.info("Trying tighter tolerance (DRMtol*0.1)") self.DRM.tol = float(self.likelihoodConf['drmtol']) * 0.1 self.DRM.fit(verbosity= int(self.commonConf['verbosity'])) except: self.logger.error("Second DRM Fit Failed") try: self.logger.info("Trying looser tolerance (drmtol*10.)") self.DRM.tol = float(self.likelihoodConf['drmtol']) * 10. self.DRM.fit(verbosity= int(self.commonConf['verbosity'])) except: self.logger.error("Third DRM Fit Failed") try: self.logger.info("Trying alternate fit algorithm (MINUIT)") self.initAltFit() self.ALTFIT.fit(verbosity=int(self.commonConf['verbosity']),covar=True,optObject=self.ALTFITobj) print self.ALTFITobj.getQuality() altfit = True except: self.logger.error("Alternative fit algorithm failed, bailing") self.logger.error(self.decodeRetCode('Minuit',self.ALTFITobj.getRetCode())) self.ALTFIT.logLike.writeXml(self.commonConf['base']+'_badDRMFit.xml') self.logger.info("Saved ALTFIT as "+self.commonConf['base']+"_badDRMFit.xml") return if(altfit): self.logger.info("ALTFIT Fit Finished. Total TS: "+str(self.ALTFIT.logLike.value())) self.ALTFIT.logLike.writeXml(self.commonConf['base']+'_likeDRM.xml') self.logger.info("Saved ALTFIT as "+self.commonConf['base']+"_likeDRM.xml") else: self.DRM.logLike.writeXml(self.commonConf['base']+'_likeDRM.xml') self.logger.info("DRM Fit Finished. Total TS: "+str(self.DRM.logLike.value())) self.logger.info("Saved DRM as "+self.commonConf['base']+"_likeDRM.xml") def fitMIN(self): """Does a New Minuit fit on your data based on the model output by the fitDRM function. You need to have run initMIN before running this function. Saves the results to <basename>_likeMIN.xml if there is convergence. If convergence is not found, saves the results to <basename>_badMINFit.xml.""" try: self.MIN except AttributeError: self.logger.critical("MIN object does not exist. Create it first with the initMIN function.") return self.MIN.fit(covar=True, optObject=self.MINobj,verbosity=int(self.commonConf['verbosity'])) self.MIN.logLike.writeXml(self.commonConf['base']+'_likeMinuit.xml') self.logger.info("NEWMINUIT Fit Finished. Total TS: "+str(self.MIN.logLike.value())) self.logger.info("NEWMINUIT Fit Status: "+str(self.MINobj.getRetCode())) self.logger.info("NEWMINUIT fit Distance: "+str(self.MINobj.getDistance())) self.fitbit = True if(self.MINobj.getRetCode() > 0): self.logger.error("NEWMINUIT DID NOT CONVERGE!!!") self.logger.error("The fit failed the following tests: "+self.decodeRetCode('NewMinuit',self.MINobj.getRetCode())) self.MIN.fit(covar=True, optObject=self.MINobj,verbosity=int(self.commonConf['verbosity'])) self.MIN.logLike.writeXml(self.commonConf['base']+'_badMINFit.xml') def printSource(self,source,Emin=100,Emax=300000): """Prints various details for a source in your model.""" try: self.MIN except AttributeError: self.logger.critical("MIN object does not exist. "+\ "Create it first with the initMIN function and then fit it with the fitMIN function.") return if(not self.fitbit): self.logger.warn("Fit isn't current, these values might not be correct. Fun fitMIN first.") logString = source TS = self.MIN.Ts(source) print "TS: ",TS logString += " TS: " + str(TS) NPred = self.MIN.NpredValue(source) print "Npred: ",NPred logString += " NPred: " + str(NPred) flux = self.MIN.flux(source,emin=Emin,emax=Emax) print "Flux: ",flux logString += "Flux: "+str(flux) if(self.fitbit): fluxErr = self.MIN.fluxError(source,emin=Emin,emax=Emax) print "Flux Error: ",fluxErr logString += "Flux Error: "+str(fluxErr) for paramName in self.MIN.model[source].funcs['Spectrum'].paramNames: paramValue = self.MIN.model[source].funcs['Spectrum'].getParam(paramName).value() print paramName,": ",paramValue logString += paramName + ": " + str(paramValue) + " " self.logger.info(logString) def customERange(self,Emin,Emax): """Sets a smaller energy range for the fitting of both the DRM and MIN optimization steps.""" try: self.DRM except AttributeError: self.logger.warn("DRM object doesn't exist. Energy range not modified.") else: self.DRM.setEnergyRange(Emin,Emax) self.logger.info("Set energy range for DRM to "+str(self.DRM.emin)+","+str(self.DRM.emax)) try: self.MIN except AttributeError: self.logger.warn("MIN object doesn't exist. Energy range not modified.") else: self.MIN.setEnergyRange(Emin,Emax) self.logger.info("Set energy range for MIN to "+str(self.MIN.emin)+","+str(self.MIN.emax)) def calcUpper(self,source,Emin=100,Emax=300000): """Calculates an upper limit for a source in your model.""" self.ul = UpperLimits(self.MIN) self.ul[source].compute(emin=Emin,emax=Emax) print self.ul[source].results self.logger.info(source+" UL: "+str(self.ul[source].results[0])) def removeWeak(self,mySource = '',tslimit=0,distlimit=0,RemoveFree=False,RemoveFixed=False): """This function has two main uses: it will print out details on all of the sources in your model and it will remove sources according to different requirements. If you just want to print out details, execute it this way: <obj>.removeWeak(<my_source>) Where <obj> is the quickLike object you're using here and <my_source> is the name of your source of interest. You can then remove some of these sources from the model if you like. For example, if you want to remove all of the fixed sources with TS values less than 1, execute it this way: <obj>.removeWeak(<my_source>,tslimit=1,RemoveFixed=True) You can mix and match any of the options. You could remove all sources (fixed and free) that are below a TS value of 3 and are 10 degrees from your source of interest by executing: <obj>.removeWeak(<my_source>,tslimit=3,distlimit=10,RemoveFree=True,RemoveFixed=True)""" try: self.MIN except AttributeError: self.logger.critical("MIN object does not exist. "+\ "Create it first with the initMIN function and then fit it with the fitMIN function.") return if(not self.fitbit): self.logger.warn("Fit isn't current, these values might not be correct. Run fitMIN first.") if(mySource == ''): mySource = self.likelihoodConf['sourcename'] for name in self.MIN.sourceNames(): remove = False distance = 0 sourceTS = self.MIN.Ts(name) if(self.MIN.model[name].src.getType() == 'Point'): distance = self.MIN._separation(self.MIN.model[mySource].src,self.MIN.model[name].src) if(self.MIN.freePars(name).size() > 0): indexFree = "Free" if( (sourceTS < tslimit) and (distance > distlimit) and RemoveFree ): remove = True else: indexFree = "Fixed" if( (sourceTS < tslimit) and (distance > distlimit) and RemoveFixed ): remove = True if( remove ): self.logger.info("Removing "+name+", TS: "+str(sourceTS)+", Frozen?: "+str(indexFree)+", Distance: "+str(distance)) self.MIN.deleteSource(name) else: self.logger.info("Retaining "+name+", TS: "+str(sourceTS)+", Frozen?: "+str(indexFree)+", Distance: "+str(distance)) def paramsAtLimit(self, limit = 0.1): """This function will print out any sources whoes parameters are close to their limits. You could use this to find sources that are having issues being fit. This function is useful when you're having trouble getting convergence from the New Minuit fit routine. The limit is in percentage difference of a bound. If one of the bounds is zero it uses the value of the parameter to check for closeness (absolute instead of percent differenct). The default is 0.1 (1%) difference for a measure of closeness.""" try: self.MIN except AttributeError: self.logger.critical("MIN object does not exist. "+\ "Create it first with the initMIN function and then fit it with the fitMIN function.") return if(not self.fitbit): self.logger.warn("Fit isn't current, these values might not be correct. Run fitMIN first.") for src in self.MIN.sourceNames(): for name in self.MIN.model[src].funcs['Spectrum'].paramNames: bounds = self.MIN.model[src].funcs['Spectrum'].getParam(name).getBounds() value = self.MIN.model[src].funcs['Spectrum'].getParam(name).value() try: distToLower = abs((value - bounds[0])/bounds[0]) except ZeroDivisionError: distToLower = abs(value) try: distToUpper = abs((value - bounds[1])/bounds[1]) except ZeroDivisionError: distToUpper = abs(value) if( distToLower < limit ): self.logger.error("The "+name+" ("+str(value)+") of "+src+" is close ("\ +str(distToLower)+") to its lower limit ("+str(bounds[0])+")") if( distToUpper < limit): self.logger.error("The "+name+" ("+str(value)+") of "+src+" is close ("\ +str(distToUpper)+") to its upper limit ("+str(bounds[1])+")") def decodeRetCode(self, optimizer, retCode): """Decodes the return codes from the Minuit and New Minuit fit functions. Used in the fitting functions in this module. You'll probably never use this function.""" if(optimizer == 'NewMinuit'): retCode -= 100 failure = "" if(retCode & 1): failure += " IsAboveMaxEdm" if(retCode & 2): failure += " HasCovariance" if(retCode & 4): failure += " HesseFailed" if(retCode & 8): failure += " HasMadePosDefCovar" if(retCode & 16): failure += " HasPosDefCovar" if(retCode & 32): failure += " HasAccurateCovar" if(retCode & 64): failure += " HasValidCovariance" if(retCode & 128): failure += " HasValidParameters" if(retCode & 256): failure += " IsValid" return failure if(optimizer == 'Minuit'): if(retCode == 0): failure = "Error matrix not calculated at all" if(retCode == 1): failure = "Diagonal approximation only, not accurate" if(retCode == 2): failure = "Full matrix, but forced positive-definite (i.e. not accurate)" if(retCode == 3): failure = "Full accurate covariance matrix (After MIGRAD, this is the indication of normal convergence.)" return failure
def runLike(likeIn,ecent,ftol,tslim,ttype,opt,rescaleAll,lastbinUL,wx): pts=[] errs=[] tsPts=[] gamma=[] Src=likeIn.source NBins=likeIn.NBins IRFs=likeIn.IRFs expCube=likeIn.expCube ft2=likeIn.ft2 bandModel=likeIn.bandModel nbins=likeIn.nbins phCorr=likeIn.phCorr flux=likeIn.ubAn.flux for i in range(0,nbins): #first, set up for a likelihood run print ' -Runnng Likelihood for band%i-' %i ev='%s_%ibins_band%i.fits' %(Src.replace(' ','_'),NBins,i) em='%s_%ibins_band%i_%s_em.fits' %(Src.replace(' ','_'),NBins,i,IRFs) band_obs=UnbinnedObs((ev),ft2,irfs=IRFs,expMap=em,expCube=expCube) band_like=UnbinnedAnalysis(band_obs,bandModel,opt) band_like.setFitTolType(ttype) stype=band_like.model.srcs[Src].spectrum().genericName() emin,emax=band_like.observation.roiCuts().getEnergyCuts() if phCorr!=1: for src in band_like.sourceNames(): par=band_like.normPar(src) par.setValue(par.getValue()*phCorr) #then set the scale factor to the center of the energy band, make sure it's frozen, and get index for prefactor while you're at it Ts=band_like.Ts freeze=band_like.freeze fit=band_like.fit DO=1 if stype=='PowerLaw': scale=getParamIndx(band_like,Src,'Scale') #this is where you have to use PowerLaw, for PowerLaw2 these parameters don't exist and this will cause problems pref=getParamIndx(band_like,Src,'Prefactor') freeze(scale) band_like[scale].setBounds(20,5e5) band_like[scale].setScale(1) band_like[scale]=1000.*ecent[i] #put center energies in units of GeV but xml files use MeV #multiplier=band_like[pref].getScale() #need to get the scale of the prefactor so the values will not be too large try: logFlux=log10(flux(Src,emin=emin,emax=emax)/(emax-emin)) except: logFlux=-14 newScale=max(int(floor(logFlux)),-14) print newScale band_like[pref].setScale(10.**newScale) multiplier=10.**newScale #cycle through the other point sources and adjust parameters of free point sources with PowerLaw2 models for src in band_like.sourceNames(): spec=band_like[src].funcs['Spectrum'] par=spec.normPar() if par.isFree()==True and band_like.model.srcs[src].spectrum().genericName()=='PowerLaw2': HIGH=getParamIndx(band_like,src,'UpperLimit') LOW=getParamIndx(band_like,src,'LowerLimit') band_like[HIGH].setBounds(20,5e5) #just in case, make sure no out of range error gets thrown band_like[LOW].setBounds(20,5e5) band_like[HIGH].setScale(1.) #just in case, make sure scale is MeV band_like[LOW].setScale(1.) band_like[HIGH]=emax band_like[LOW]=emin freeze(HIGH) #just in case, make sure these aren't fit values freeze(LOW) if rescaleAll==True: try: logFlux=log10(flux(src,emin=emin,emax=emax)) except: logFlux=-14 newScale=max(int(floor(logFlux)),-14) par.setScale(10**newScale) if rescaleAll==True and src!=Src and par.isFree()==True and band_like.model.srcs[src].spectrum().genericName()=='PowerLaw': try: logFlux=log10(flux(src,emin=emin,emax=emax)/(emax-emin)) except: logFlux=-14 newScale=max(int(floor(logFlux)),-14) par.setScale(10**newScale) try: fit(tol=1,verbosity=0,optimizer=opt) fail=0 except: try: fit(tol=1*10,verbosity=0,optimizer=opt) fail=0 except: try: fit(tol=1./10,verbosity=0,optimizer=opt) fail=0 except: print "Fit with optimizer %s with tolerances ~1 to look for negative or zero TS sources failed, if error bars are unrealistically small you may need to redo the fit for energy band %i manually" %(opt,i) fail=1 if fail==0: for src in band_like.sourceNames(): if src!=Src: par=band_like.normPar(src) if band_like[src].type=='PointSource' and Ts(src)<=0 and par.isFree()==True: band_like.deleteSource(src) print " -Removing %s from the model" %src scale=getParamIndx(band_like,Src,'Scale') pref=getParamIndx(band_like,Src,'Prefactor') #do the actual fit try: fit(tol=ftol,verbosity=0) except: try: print 'Trying lower tolerance of %s for band%i.' %(ftol/10,i) fit(tol=ftol/10,verbosity=0) except: try: print 'Trying higher tolerance of %s for band%i.' %(ftol*10,i) fit(tol=ftol*10,verbosity=0) except: print 'No convergence for band%i, skipping.' %i pts+=[0] errs+=[0] tsPts+=[0] pass #get the prefactor, error, and ts values val=band_like[pref].value() err=band_like[pref].error() TS=Ts(Src) if(TS<tslim or (i==(nbins-1) and lastbinUL==True)): #calculate 95% upperlimit if source TS<tslim, 25 by default (corresponds to ~5sigma) try: freeze(pref) ul=UpperLimits(band_like) UL=ul[Src].compute(emin=emin,emax=emax) val=UL[1] err=0 print ' NOTE: Band%i,' %i,'with center energy',ecent[i],'GeV, quoting 95% upper limit on flux.' except: try: print ' Tyring higher tolerance of %s for band %i to get good starting point for upper limit calculations.' %(ftol*10,i) band_like[pref].setFree(1) fit(tol=ftol*10,verbosity=0) TS=Ts(Src) if(TS<tslim or (i==(nbins-1) and lastbinUL==True)): freeze(pref) ul=UpperLimits(band_like) UL=ul[Src].compute(emin=emin,emax=emax) val=UL[1] err=0 print ' NOTE: Band%i,' %i,'with center energy',ecent[i],'GeV, quoting 95% upper limit on flux.' else: val=band_like[pref].value() err=band_like[pref].error() except: try: print ' Tyring lower tolerance of %s for band %i to get good starting point for upper limit calculations.' %(ftol/10,i) band_like[pref].setFree(1) fit(tol=ftol/10,verbosity=0) TS=Ts(Src) if(TS<tslim or (i==(nbins-1) and lastbinUL==True)): freeze(pref) ul=UpperLimits(band_like) UL=ul[Src].compute(emin=emin,emax=emax) val=UL[1] err=0 print ' NOTE: Band%i,' %i,'with center energy',ecent[i],'GeV, quoting 95% upper limit on flux.' else: val=band_like[pref].value() err=band_like[pref].error() except: err=0 print ' NOTE: Band%i,' %i,'with center energy',ecent[i],'GeV, TS<%s' %tslim,'but UpperLimits computation failed.' print ' Quoting best fit value with zero error.' tsPts+=[TS] pts+=[val*multiplier/phCorr] errs+=[err*multiplier/phCorr] elif stype=='PowerLaw2': Upper=getParamIndx(band_like,Src,'UpperLimit') Lower=getParamIndx(band_like,Src,'LowerLimit') Integral=getParamIndx(band_like,Src,'Integral') Index=getParamIndx(band_like,Src,'Index') freeze(Upper) freeze(Lower) band_like[Upper].setBounds(20,5e5) band_like[Lower].setBounds(20,5e5) band_like[Lower].setScale(1) band_like[Lower].setScale(1) band_like[Upper]=emax band_like[Lower]=emin #multiplier=band_like[Integral].getScale() try: logFlux=log10(flux(Src,emin=emin,emax=emax)) except: logFlux=-14 newScale=max(int(floor(logFlux)),-14) band_like[Integral].setScale(10**newScale) multiplier=10**newScale indxMult=band_like[Index].getScale() for src in band_like.sourceNames(): if src!=Src: spec=band_like[src].funcs['Spectrum'] par=spec.normPar() if par.isFree()==True and band_like.model.srcs[src].spectrum().genericName()=='PowerLaw2': HIGH=getParamIndx(band_like,src,'UpperLimit') LOW=getParamIndx(band_like,src,'LowerLimit') band_like[HIGH].setBounds(20,5e5) #just in case, make sure no out of range error gets thrown band_like[LOW].setBounds(20,5e5) band_like[HIGH].setScale(1.) #just in case, make sure scale is MeV band_like[LOW].setScale(1.) band_like[HIGH]=emax band_like[LOW]=emin freeze(HIGH) #just in case, make sure these aren't fit values freeze(LOW) if rescaleAll==True: try: logFlux=log10(flux(src,emin=emin,emax=emax)) except: logFlux=-14 newScale=max(int(floor(logFlux)),-14) par.setScale(10**newScale) if rescaleAll==True and src!=Src and par.isFree()==True and band_like.model.srcs[src].spectrum().genericName()=='PowerLaw': try: logFlux=log10(flux(src,emin=emin,emax=emax)/(emax-emin)) except: logFlux=-14 newScale=max(int(floor(logFlux)),-14) par.setScale(10**newScale) try: fit(tol=1,verbosity=0,optimizer=opt) fail=0 except: try: fit(tol=1*10,verbosity=0,optimizer=opt) fail=0 except: try: fit(tol=1./10,verbosity=0,optimizer=opt) fail=0 except: print "Fit with optimizer %s with tolerances ~1 to look for negative or zero TS sources failed, if error bars are unrealistically small you may need to redo the fit for energy band %i manually" %(opt,i) fail=1 if not fail: for src in band_like.sourceNames(): if src!=Src: par=band_like.normPar(src) if band_like[src].type=='PointSource' and Ts(src)<=0 and par.isFree()==True: band_like.deleteSource(src) print " -Removing %s from the model" %src Upper=getParamIndx(band_like,Src,'UpperLimit') Lower=getParamIndx(band_like,Src,'LowerLimit') Integral=getParamIndx(band_like,Src,'Integral') Index=getParamIndx(band_like,Src,'Index') try: fit(tol=ftol,verbosity=0) except: try: print 'Trying lower tolerance of %s for band%i.' %(ftol/10,i) fit(tol=ftol/10,verbosity=0) except: try: print 'Trying higher tolerance of %s for band%i.' %(ftol*10,i) fit(tol=ftol*10,verbosity=0) except: print 'No convergence for band%i, skipping.' %i pts+=[0] errs+=[0] tsPts+=[0] gamma+=[0] DO=0 if DO: val=band_like[Integral].value() err=band_like[Integral].error() TS=Ts(Src) gam=band_like[Index].value()*indxMult*-1. if(TS<tslim or (i==(nbins-1) and lastbinUL==True)): #calculate 95% upperlimit if source TS<9, i.e. less than 3 sigma detection in each energy band try: freeze(Integral) ul=UpperLimits(band_like) UL=ul[Src].compute(emin=emin,emax=emax) val=UL[1] err=0 #need to redo the fit with band_like set to Upper Limit value to get correct spectral index for that value band_like[Integral]=val freeze(Integral) fit(tol=ftol,verbosity=0) gam=band_like[Index].value()*indxMult*-1. print ' NOTE: Band%i,' %i,'with center energy',ecent[i],'GeV, quoting 95% upper limit on flux.' except: try: print ' Tyring higher tolerance of %s for band %i to get good starting point for upper limit calculations.' %(ftol*10,i) band_like[Integral].setFree(1) fit(tol=ftol*10,verbosity=0) TS=Ts(Src) if(TS<tslim or (i==(nbins-1) and lastbinUL==True)): freeze(Integral) ul=UpperLimits(band_like) UL=ul[Src].compute(emin=emin,emax=emax) val=UL[1] err=0 band_like[Integral]=val freeze(Integral) fit(tol=ftol*10,verbosity=0) gam=band_like[Index].value()*indxMult*-1. print ' NOTE: Band%i,' %i,'with center energy',ecent[i],'GeV, quoting 95% upper limit on flux.' else: val=band_like[Integral].value() err=band_like[Integral].error() gam=band_like[Index].value()*indxMult*-1. except: try: print ' Tyring lower tolerance of %s for band %i to get good starting point for upper limit calculations.' %(ftol/10,i) band_like[Integral].setFree(1) fit(tol=ftol/10,verbosity=0) TS=Ts(Src) if(TS<tslim or (i==(nbins-1) and lastbinUL==True)): freeze(Integral) ul=UpperLimits(band_like) UL=ul[Src].compute(emin=emin,emax=emax) val=UL[1] err=0 band_like[Integral]=val freeze(Integral) fit(tol=ftol/10,verbosity=0) gam=band_like[Index].value()*indxMult*-1. print ' NOTE: Band%i,' %i,'with center energy',ecent[i],'GeV, quoting 95% upper limit on flux.' else: val=band_like[Integral].value() err=band_like[Integral].error() gam=band_like[Index].value()*indxMult*-1. except: err=0 print ' NOTE: Band%i,' %i,'with center energy',ecent[i],'GeV, TS<%s' %tslim,'but UpperLimits computation failed.' print ' Quoting best fit value with zero error.' tsPts+=[TS] pts+=[val*multiplier/phCorr] errs+=[err*multiplier/phCorr] gamma+=[gam] else: print '%s needs to have PowerLaw or PowerLaw2 spectral model, not %s' %(Src,stype) print 'exiting without running likelihood in the energy bands' return None,None,None,None if wx: band_like.writeXml('%s_%ibins_band%i_fitmodel.xml'%(Src.replace(' ','_'),NBins,i)) del band_like del band_obs return pts,errs,tsPts,gamma