def calcCompleteness(orbitfile, obsfile, outDir):
"""
Calculate a basic set of metrics on the NEOs, including completeness.
Saves the plots to disk.
"""
# Read data back from disk.
mos = MoSlicer(orbitfile, Hrange=np.arange(13, 26, 0.5))
mos.readObs(obsfile)
# Nobs
metric = MoMetrics.NObsMetric()
slicer = mos
pandasConstraint = None
plotDict = {'nxbins':100, 'nybins':100}
nobs = mmb.MoMetricBundle(metric, slicer, pandasConstraint,
runName=runName, metadata=metadata, plotDict=plotDict)
# Calculate completeness. First we must calculate "DiscoveryChances".
# Set up an example metric bundle.
metric = MoMetrics.DiscoveryChancesMetric()
slicer = mos
pandasConstraint = None
discovery = mmb.MoMetricBundle(metric, slicer, pandasConstraint,
runName=runName, metadata=metadata, plotDict=plotDict)
bdict = {'nobs':nobs, 'discovery':discovery}
bg = mmb.MoMetricBundleGroup(bdict, outDir=outDir)
bg.runAll()
bg.plotAll()
ph = plots.PlotHandler(outDir=outDir)
# Then calculate 'completeness' as function of H, as a secondary metric.
completeness = discovery.reduceMetric(discovery.metric.reduceFuncs['Completeness'])
# And we can make an 'integrated over H distribution' version.
completenessInt = completeness.reduceMetric(completeness.metric.reduceFuncs['CumulativeH'])
Hmark = 21.0
for c in [completeness, completenessInt]:
summaryMetric = ValueAtHMetric(Hmark=Hmark)
c.setSummaryMetrics(summaryMetric)
c.computeSummaryStats()
label = "Completeness at H=%.1f: %.2f" %(Hmark, c.summaryValues['Value At H=%.1f' %Hmark])
c.setPlotDict({'label':label})
c.plot(plotFunc = moPlots.MetricVsH())
plt.axvline(Hmark, color='r', linestyle=':')
plt.axhline(c.summaryValues['Value At H=%.1f' %(Hmark)], color='r', linestyle='-')
plt.legend(loc=(0.9, 0.2))
plt.savefig(os.path.join(outDir, c.fileRoot + '_vsH' + '.png'), format='png')
def reduceMetric(self,
reduceFunc,
reducePlotDict=None,
reduceDisplayDict=None):
"""
Run reduce methods on the metric bundle, such as completeness or integrate over H distribution.
These return a new metric bundle.
"""
rName = reduceFunc.__name__.replace('reduce', '')
reduceName = self.metric.name + '_' + rName
newmetric = deepcopy(self.metric)
newmetric.name = reduceName
newBundle = MoMetricBundle(metric=newmetric,
slicer=self.slicer,
constraint=self.constraint,
runName=self.runName,
metadata=self.metadata,
plotDict=self.plotDict,
plotFuncs=self.plotFuncs,
summaryMetrics=self.summaryMetrics)
if rName == 'Completeness':
newBundle.metric.name = 'Completeness'
newBundle.plotFuncs = [moPlots.MetricVsH()]
newBundle.setPlotDict({'ylabel': 'Completeness @H'})
if rName == 'CumulativeH':
newBundle.metric.name = self.metric.name + ' (cumulative H)'
newBundle.setPlotDict({'units': '<=H'})
if 'ylabel' in newBundle.plotDict:
newBundle.plotDict['ylabel'] = newBundle.plotDict[
'ylabel'].replace('@H', '<=H')
newBundle._buildFileRoot()
# Calculate new metric values.
newBundle.metricValues, Hvals = reduceFunc(
self.metricValues, self.slicer.slicePoints['H'])
if newBundle.metricValues.shape[1] != self.slicer.slicerShape[1]:
# Then something happened with reduceFunction --
# .. usually, this would be with 'completeness'. It can reshape the metric values so that
# (a) if we cloned H, we now go from multiple metricvalues per H value to a single value per H [(nSso, nHrange) -> (1, nHrange)]
# (b) if we did not clone H, we now have a binned metricvalues - one per H bin, instead of one per nsso [(nSso, 1) -> (nHrange, nHrange)]
# and we don't really care (in general) because we won't be re-slicing the observations. But we do need to update the slicePoints['H'],
# and for that we need a new slicer. Rereading the orbit file is a bit of a pain, but not too bad.
newslicer = MoSlicer(orbitfile=self.slicer.orbitfile, Hrange=Hvals)
newBundle.slicer = newslicer
return newBundle
def calcCompleteness(orbitfile, obsfile, outDir, runName, metadata=None):
"""
Calculate a basic set of metrics on the NEOs, including completeness.
Saves the plots to disk.
"""
# Read data back from disk.
mos = MoSlicer(orbitfile, Hrange=np.arange(15, 27, 0.25))
mos.readObs(obsfile)
# Nobs
metric = moMetrics.NObsMetric()
slicer = mos
pandasConstraint = None
plotDict = {'nxbins': 200, 'nybins': 200}
nobs = mmb.MoMetricBundle(metric,
slicer,
pandasConstraint,
runName=runName,
metadata=metadata,
plotDict=plotDict)
# Calculate completeness. First we must calculate "DiscoveryChances".
# Set up an example metric bundle.
discovery = {}
nObsList = [2, 3, 4]
nyears = [3, 5, 10, 12, 15]
for yr in nyears:
discovery[yr] = {}
for nObs in nObsList:
md = metadata + ' %d visits/night after %d years' % (nObs, yr)
#metric = moMetrics.DiscoveryMetric(tMin=0, tMax=90./60/24., nObsPerNight=nObs, nNightsPerWindow=3, tWindow=30)
metric = moMetrics.DiscoveryChancesMetric(tNight=90. / 60. / 24.,
nObsPerNight=nObs,
nNightsPerWindow=3,
tWindow=15)
slicer = mos
pandasConstraint = 'night <= %d' % (yr * 365)
discovery[yr][nObs] = mmb.MoMetricBundle(
metric,
slicer,
pandasConstraint,
runName=runName,
metadata=md,
plotDict=plotDict,
plotFuncs=[moPlots.MetricVsH()])
#metric = moMetrics.DiscoveryMetric(tMin=0, tMax=90./60/24., nObsPerNight=2, nNightsPerWindow=4, tWindow=30)
metric = moMetrics.DiscoveryChancesMetric(tNight=90. / 60. / 24.,
nObsPerNight=2,
nNightsPerWindow=4,
tWindow=30)
discovery[yr]['4nights'] = mmb.MoMetricBundle(
metric=metric,
slicer=mos,
constraint=pandasConstraint,
runName=runName,
metadata=metadata + '4 nights/track after %d years' % (yr),
plotDict=plotDict,
plotFuncs=[moPlots.MetricVsH()])
bdict = {}
for yr in nyears:
for nObs in nObsList:
bdict['discovery_%s_%d' % (nObs, yr)] = discovery[yr][nObs]
bdict['4nights_%d' % (yr)] = discovery[yr]['4nights']
bdict['nobs'] = nobs
bg = mmb.MoMetricBundleGroup(bdict, outDir=outDir)
bg.runAll()
bg.plotAll()
Hmark = 22
completeness = {}
completenessInt = {}
hVals = mos.Hrange
for yr in nyears:
completeness[yr] = {}
completenessInt[yr] = {}
for nObs in nObsList:
#discChances = discovery[nObs].childBundles['N_Chances']
discChances = discovery[yr][nObs]
discChances.setSummaryMetrics([
moSummaryMetrics.CompletenessMetric(),
moSummaryMetrics.CumulativeCompletenessMetric()
])
discChances.computeSummaryStats()
completeness[yr][nObs] = discChances.summaryValues['Completeness'][
0]
completenessInt[yr][nObs] = discChances.summaryValues[
'CumulativeCompleteness'][0]
for nObs in ['4nights']:
#discChances = discovery[nObs].childBundles['N_Chances']
discChances = discovery[yr][nObs]
discChances.setSummaryMetrics([
moSummaryMetrics.CompletenessMetric(),
moSummaryMetrics.CumulativeCompletenessMetric()
])
discChances.computeSummaryStats()
completeness[yr][nObs] = discChances.summaryValues['Completeness'][
0]
completenessInt[yr][nObs] = discChances.summaryValues[
'CumulativeCompleteness'][0]
# Make a figure of completeness with the different discovery patterns, for one year.
yr = 12
Hidx = np.where(hVals == Hmark)[0]
plt.figure()
plt.title('Completeness %s' % (runName))
for nObs in nObsList:
cval = completeness[yr][nObs][Hidx]
plt.plot(hVals,
completeness[yr][nObs],
label='%d obs/tracklet, %.0f%s H @ %.1f' %
(nObs, cval * 100, '%', Hmark))
for nObs in ['4nights']:
cval = completeness[yr][nObs][Hidx]
plt.plot(hVals,
completeness[yr][nObs],
label='4 pairs/track, %.0f%s H @ %.1f' %
(cval * 100, '%', Hmark))
plt.axvline(Hmark, color='r', linestyle=':')
plt.xlabel('H')
plt.ylabel('Completeness @ H')
plt.legend(loc='upper right',
fancybox=True,
numpoints=1,
fontsize='smaller')
plt.savefig(os.path.join(outDir, 'completeness.pdf'), format='pdf')
plt.figure()
plt.title('Cumulative completeness %s' % (runName))
for nObs in nObsList:
cval = completenessInt[yr][nObs][Hidx]
plt.plot(hVals,
completenessInt[yr][nObs],
label='%d obs/tracklet, %.0f%s H <= %.1f' %
(nObs, cval * 100, '%', Hmark))
for nObs in ['4nights']:
cval = completenessInt[yr][nObs][Hidx]
plt.plot(hVals,
completenessInt[yr][nObs],
label='4 pairs/track, %.0f%s H <= %.1f' %
(cval * 100, '%', Hmark))
plt.axvline(Hmark, color='r', linestyle=':')
plt.xlabel('H')
plt.ylabel('Completeness <= H')
plt.legend(loc='upper right',
fancybox=True,
numpoints=1,
fontsize='smaller')
plt.savefig(os.path.join(outDir, 'completenessInt.pdf'), format='pdf')
# And make a figure for one set of discovery criterion, over time.
nObs = 2
plt.figure()
plt.title('Completeness over time %s' % (runName))
for yr in nyears:
cval = completeness[yr][nObs][Hidx]
plt.plot(hVals,
completeness[yr][nObs],
label='Year %d: %.0f @ H=%.1f' % (yr, cval * 100, Hmark))
plt.axvline(Hmark, color='r', linestyle=':')
plt.xlabel('H')
plt.ylabel('Completeness @ H')
plt.legend(loc='upper right',
fancybox=True,
numpoints=1,
fontsize='smaller')
plt.savefig(os.path.join(outDir, 'completenessTime.pdf'), format='pdf')
nObs = 2
plt.figure()
plt.title('Completeness over time %s' % (runName))
for yr in nyears:
cval = completenessInt[yr][nObs][Hidx]
plt.plot(hVals,
completenessInt[yr][nObs],
label='Year %d: %.0f <= H=%.1f' % (yr, cval * 100, Hmark))
plt.axvline(Hmark, color='r', linestyle=':')
plt.xlabel('H')
plt.ylabel('Completeness <= H')
plt.legend(loc='upper right',
fancybox=True,
numpoints=1,
fontsize='smaller')
plt.savefig(os.path.join(outDir, 'completenessIntTime.pdf'), format='pdf')
def calcCompleteness(orbitfile, obsfile, outDir, runName, metadata=None):
"""
Calculate a basic set of metrics on the NEOs, including completeness.
Saves the plots to disk.
"""
# Read data back from disk.
mos = MoSlicer(orbitfile, Hrange=np.arange(15, 27, 0.25))
mos.readObs(obsfile)
# Nobs
metric = moMetrics.NObsMetric()
slicer = mos
pandasConstraint = None
plotDict = {'nxbins':200, 'nybins':200}
nobs = mmb.MoMetricBundle(metric, slicer, pandasConstraint,
runName=runName, metadata=metadata, plotDict=plotDict)
# Calculate completeness. First we must calculate "DiscoveryChances".
# Set up an example metric bundle.
discovery = {}
nObsList = [2, 3, 4]
nyears = [3, 5, 10, 12, 15]
for yr in nyears:
discovery[yr] = {}
for nObs in nObsList:
md = metadata + ' %d visits/night after %d years' %(nObs, yr)
#metric = moMetrics.DiscoveryMetric(tMin=0, tMax=90./60/24., nObsPerNight=nObs, nNightsPerWindow=3, tWindow=30)
metric = moMetrics.DiscoveryChancesMetric(tNight=90./60./24., nObsPerNight=nObs, nNightsPerWindow=3, tWindow=15)
slicer = mos
pandasConstraint = 'night <= %d' %(yr*365)
discovery[yr][nObs] = mmb.MoMetricBundle(metric, slicer, pandasConstraint,
runName=runName, metadata=md,
plotDict=plotDict, plotFuncs=[moPlots.MetricVsH()])
#metric = moMetrics.DiscoveryMetric(tMin=0, tMax=90./60/24., nObsPerNight=2, nNightsPerWindow=4, tWindow=30)
metric = moMetrics.DiscoveryChancesMetric(tNight=90./60./24., nObsPerNight=2, nNightsPerWindow=4, tWindow=30)
discovery[yr]['4nights'] = mmb.MoMetricBundle(metric=metric,
slicer=mos, constraint=pandasConstraint, runName=runName,
metadata = metadata+'4 nights/track after %d years' %(yr),
plotDict=plotDict, plotFuncs=[moPlots.MetricVsH()])
bdict = {}
for yr in nyears:
for nObs in nObsList:
bdict['discovery_%s_%d' %(nObs, yr)] = discovery[yr][nObs]
bdict['4nights_%d' %(yr)] = discovery[yr]['4nights']
bdict['nobs'] = nobs
bg = mmb.MoMetricBundleGroup(bdict, outDir=outDir)
bg.runAll()
bg.plotAll()
Hmark = 22
completeness = {}
completenessInt = {}
hVals = mos.Hrange
for yr in nyears:
completeness[yr] = {}
completenessInt[yr] = {}
for nObs in nObsList:
#discChances = discovery[nObs].childBundles['N_Chances']
discChances = discovery[yr][nObs]
discChances.setSummaryMetrics([moSummaryMetrics.CompletenessMetric(), moSummaryMetrics.CumulativeCompletenessMetric()])
discChances.computeSummaryStats()
completeness[yr][nObs] = discChances.summaryValues['Completeness'][0]
completenessInt[yr][nObs] = discChances.summaryValues['CumulativeCompleteness'][0]
for nObs in ['4nights']:
#discChances = discovery[nObs].childBundles['N_Chances']
discChances = discovery[yr][nObs]
discChances.setSummaryMetrics([moSummaryMetrics.CompletenessMetric(), moSummaryMetrics.CumulativeCompletenessMetric()])
discChances.computeSummaryStats()
completeness[yr][nObs] = discChances.summaryValues['Completeness'][0]
completenessInt[yr][nObs] = discChances.summaryValues['CumulativeCompleteness'][0]
# Make a figure of completeness with the different discovery patterns, for one year.
yr = 12
Hidx = np.where(hVals == Hmark)[0]
plt.figure()
plt.title('Completeness %s' %(runName))
for nObs in nObsList:
cval = completeness[yr][nObs][Hidx]
plt.plot(hVals, completeness[yr][nObs], label='%d obs/tracklet, %.0f%s H @ %.1f' %(nObs, cval*100, '%', Hmark))
for nObs in ['4nights']:
cval = completeness[yr][nObs][Hidx]
plt.plot(hVals, completeness[yr][nObs], label='4 pairs/track, %.0f%s H @ %.1f' %(cval*100, '%', Hmark))
plt.axvline(Hmark, color='r', linestyle=':')
plt.xlabel('H')
plt.ylabel('Completeness @ H')
plt.legend(loc='upper right', fancybox=True, numpoints=1, fontsize='smaller')
plt.savefig(os.path.join(outDir, 'completeness.pdf'), format='pdf')
plt.figure()
plt.title('Cumulative completeness %s' %(runName))
for nObs in nObsList:
cval = completenessInt[yr][nObs][Hidx]
plt.plot(hVals, completenessInt[yr][nObs], label='%d obs/tracklet, %.0f%s H <= %.1f' %(nObs, cval*100, '%', Hmark))
for nObs in ['4nights']:
cval = completenessInt[yr][nObs][Hidx]
plt.plot(hVals, completenessInt[yr][nObs], label='4 pairs/track, %.0f%s H <= %.1f' %(cval*100, '%', Hmark))
plt.axvline(Hmark, color='r', linestyle=':')
plt.xlabel('H')
plt.ylabel('Completeness <= H')
plt.legend(loc='upper right', fancybox=True, numpoints=1, fontsize='smaller')
plt.savefig(os.path.join(outDir, 'completenessInt.pdf'), format='pdf')
# And make a figure for one set of discovery criterion, over time.
nObs = 2
plt.figure()
plt.title('Completeness over time %s' %(runName))
for yr in nyears:
cval = completeness[yr][nObs][Hidx]
plt.plot(hVals, completeness[yr][nObs], label='Year %d: %.0f @ H=%.1f' %(yr, cval*100, Hmark))
plt.axvline(Hmark, color='r', linestyle=':')
plt.xlabel('H')
plt.ylabel('Completeness @ H')
plt.legend(loc='upper right', fancybox=True, numpoints=1, fontsize='smaller')
plt.savefig(os.path.join(outDir, 'completenessTime.pdf'), format='pdf')
nObs = 2
plt.figure()
plt.title('Completeness over time %s' %(runName))
for yr in nyears:
cval = completenessInt[yr][nObs][Hidx]
plt.plot(hVals, completenessInt[yr][nObs], label='Year %d: %.0f <= H=%.1f' %(yr, cval*100, Hmark))
plt.axvline(Hmark, color='r', linestyle=':')
plt.xlabel('H')
plt.ylabel('Completeness <= H')
plt.legend(loc='upper right', fancybox=True, numpoints=1, fontsize='smaller')
plt.savefig(os.path.join(outDir, 'completenessIntTime.pdf'), format='pdf')