def getHist(data,var_name,mask=None,edges=None):
h = Histogram(edges)
if not mask is None:
h.fill(data[var_name][mask])
else:
h.fill(data[var_name])
return h
def bootstrap(self,fraction=0.68,niter=100,xmax=None):
if xmax is None:
nedge = len(self._xedge)
else:
nedge = len(self._ncounts[self._xedge<=xmax])
xedge = self._xedge[:nedge]
h = Histogram.createFromCDF(xedge,self._ncounts[1:nedge])
nbkg = np.random.poisson(self._nbkg,niter).astype(float)
ncounts = np.random.poisson(np.concatenate(([0],h.counts)),
(niter,nedge))
xq = []
for i in range(niter):
x = self._quantile(nbkg[i],self._xedge[:nedge],
ncounts[i],fraction)
xq.append(x)
xq_mean = np.mean(np.array(xq))
xq_rms = np.std(np.array(xq))
return xq_mean, xq_rms
def fillLivetime(self, ltc, src_type):
self._tau = Histogram(self._cth_axis)
if src_type == 'iso':
self._tau += self._cth_axis.width
elif src_type == 'isodec':
sinlat = np.linspace(-1, 1, 48)
m = ltc._ltmap[:, i]
self._tau[i] = 0
for s in sinlat:
lat = np.arcsin(s)
th = np.pi / 2. - lat
ipix = healpy.ang2pix(64, th, 0, nest=True)
self._tau[i] += m[ipix]
else:
if not isinstance(src_type, list):
src_type = [src_type]
for s in src_type:
cat = Catalog.get()
src = cat.get_source_by_name(s)
lonlat = (src['RAJ2000'], src['DEJ2000'])
self._tau += ltc.get_src_lthist(lonlat[0],
lonlat[1],
cth_axis=self._cth_axis)
def bootstrap(self,fraction=0.68,niter=100,xmax=None):
nedge = len(self._non)
hon = Histogram.createFromCDF(self._axis.edges,self._non[1:])
hoff = Histogram.createFromCDF(self._axis.edges,self._noff[1:])
non = np.random.poisson(np.concatenate(([0],hon.counts)),
(niter,nedge))
noff = np.random.poisson(np.concatenate(([0],hoff.counts)),
(niter,nedge))
xq = []
for i in range(niter):
xq.append(self._quantile(non[i],noff[i],fraction))
xq_mean = np.mean(np.array(xq))
xq_rms = np.std(np.array(xq))
return xq_mean, xq_rms
def get_spectrum(self,lon,lat):
xy = self._wcs.wcs_world2pix(lon,lat, 0)
ilon = np.round(xy[0][0])
ilat = np.round(xy[1][0])
ilon = min(max(0,ilon),self._axes[0]._naxis-1)
ilat = min(max(0,ilat),self._axes[1]._naxis-1)
c = self._counts.T[ilon,ilat,:]
edges = self._axes[2].edges
return Histogram.createFromArray(edges,c)
def getHist(data, var_name, mask=None, edges=None):
h = Histogram(edges)
if not mask is None:
h.fill(data[var_name][mask])
else:
h.fill(data[var_name])
return h
def getOnOffHist(data, var_name, phases, mask=None, edges=None):
(on_phase, off_phase, alpha) = phases
on_mask = PhotonData.get_mask(data, phases=on_phase)
off_mask = PhotonData.get_mask(data, phases=off_phase)
if not mask is None:
on_mask &= mask
off_mask &= mask
hon = Histogram(edges)
hon.fill(data[var_name][on_mask])
hoff = Histogram(edges)
hoff.fill(data[var_name][off_mask])
hoffs = copy.deepcopy(hoff)
hoffs *= alpha
return (hon, hoff, hoffs)
def __init__(self,
irf,
nbin=600,
ebins_per_decade=16,
src_type='iso',
spectrum='powerlaw',
spectrum_pars=[2.0],
ltcube=None,
edisp_table=None):
super(PSFModelLT, self).__init__(model=spectrum,
sp_param=spectrum_pars)
self._nbin_dtheta = nbin
self._irf = irf
self._edisp_table = edisp_table
loge_step = 1. / float(ebins_per_decade)
emin = 1.0 + loge_step / 2.
emax = 6.0 - loge_step / 2.
nbin = int((emax - emin) / loge_step) + 1
self._loge_axis = Axis.create(emin, emax, nbin)
self._energy = np.power(10, self._loge_axis.center)
self._exps = np.zeros(self._loge_axis.nbins)
self._psf = np.zeros((self._loge_axis.nbins, self._nbin_dtheta))
self._dtheta = np.array([
self._dtheta_max_deg * (float(i) / float(self._nbin_dtheta))**2
for i in range(self._nbin_dtheta)
])
self._dtheta_axis = Axis(self._dtheta)
self._cth_axis = Axis.create(0.2, 1.0, 40)
self._tau = Histogram(self._cth_axis)
self.fillLivetime(ltcube, src_type)
self.buildModel()
def getOnOffHist(data,var_name,phases,mask=None,edges=None):
(on_phase,off_phase,alpha) = phases
on_mask = PhotonData.get_mask(data,phases=on_phase)
off_mask = PhotonData.get_mask(data,phases=off_phase)
if not mask is None:
on_mask &= mask
off_mask &= mask
hon = Histogram(edges)
hon.fill(data[var_name][on_mask])
hoff = Histogram(edges)
hoff.fill(data[var_name][off_mask])
hoffs = copy.deepcopy(hoff)
hoffs *= alpha
return (hon,hoff,hoffs)
def histogram(self, emin, emax, cthmin, cthmax, edges):
y = self.thetasq(emin, emax, cthmin, cthmax, edges)
return Histogram(edges, counts=y)
print 'SIGMA ', sigma
print 'SIGMA ', sigma_mu_fn(1.0)
print 'SIGMA ', sigma_mu_fn(10.0)
print 'Asimov q0 UL ', find_fn_root(lambda t: fn_qmu(mub+t,mub+t,mub),0,100,dlnl)
print 'Asimov qmu UL ', ul_asimov_qmu, sigma
print 'Asimov UL ', ul_asimov, ul_asimov_upper-ul_asimov, ul_asimov_lower-ul_asimov
print -2*poisson_delta_lnl(mub,mub,mub+ul_asimov)
print -2*poisson_delta_lnl(mub,mub,mub+ul_asimov_qmu)
print fn_qmu(mub,mub,mub+ul_asimov_qmu)
print dlnl
qmu = -2*poisson_delta_lnl(mub,mub,mub+ul_asimov_qmu)
h = Histogram(Axis.create(0,100,100))
h.fill(ul_mc)
h.normalize().cumulative().plot()
plt.axhline(0.5-0.34,color='g')
plt.axhline(0.5+0.34,color='g')
plt.axvline(ul_asimov_qmu,color='k')
plt.axvline(ul_asimov_qmu+sigma,color='k',linestyle='--')
plt.axvline(ul_asimov_qmu-sigma,color='k',linestyle='--')
# plt.axvline(ul_asimov_qmu-sigma/(ul_asimov_qmu+sigma)*ul_asimov_qmu,
# color='k',linestyle='--')
def make_image_plot(self,subplot,h,fig,fig2,title,rpsf68,rpsf95,
smooth=False,
resid_type=None,mc_resid=None,**kwargs):
plt.figure(fig.get_label())
cb_label='Counts'
if resid_type == 'significance':
kwargs['vmin'] = -5
kwargs['vmax'] = 5
kwargs['levels'] = [-5.0,-3.0,3.0,5.0]
cb_label = 'Significance [$\sigma$]'
elif resid_type == 'fractional':
kwargs['vmin'] = -1.0
kwargs['vmax'] = 1.0
kwargs['levels'] = [-1.0,-0.5,0.5,1.0]
cb_label = 'Fractional Residual'
if smooth:
kwargs['beam_size'] = [self._rsmooth,self._rsmooth,0.0,4]
axim = h.plot(subplot=subplot,cmap='ds9_b',**kwargs)
h.plot_circle(rpsf68,color='w',lw=1.5)
h.plot_circle(rpsf95,color='w',linestyle='--',lw=1.5)
h.plot_marker(marker='+',color='w',linestyle='--')
ax = h.ax()
ax.set_title(title)
cb = plt.colorbar(axim,orientation='horizontal',
shrink=0.9,pad=0.15,
fraction=0.05)
cb.set_label(cb_label)
cat = Catalog.get('3fgl')
cat.plot(h,ax=ax,src_color='w',label_threshold=5.0)
if resid_type is None: return
plt.figure(fig2.get_label())
ax2 = fig2.add_subplot(subplot)
z = h.counts[10:-10,10:-10]
if resid_type == 'significance':
zproj_axis = Axis.create(-6,6,120)
elif resid_type == 'fractional':
zproj_axis = Axis.create(-1.0,1.0,120)
else:
zproj_axis = Axis.create(-10,10,120)
hz = Histogram(zproj_axis)
hz.fill(np.ravel(z))
nbin = np.prod(z.shape)
hz_mc = Histogram(zproj_axis)
if mc_resid:
for mch in mc_resid:
z = mch.counts[10:-10,10:-10]
hz_mc.fill(np.ravel(z))
hz_mc /= float(len(mc_resid))
fn = lambda t : 1./np.sqrt(2*np.pi)*np.exp(-t**2/2.)
hz.plot(label='Data',linestyle='None')
if resid_type == 'significance':
plt.plot(hz.axis().center,
fn(hz.axis().center)*hz.axis().width*nbin,
color='k',label='Gaussian ($\sigma = 1$)')
hz_mc.plot(label='MC',hist_style='line')
plt.gca().grid(True)
plt.gca().set_yscale('log')
plt.gca().set_ylim(0.5)
ax2.legend(loc='upper right',prop= {'size' : 10 })
data_stats = 'Mean = %.2f\nRMS = %.2f'%(hz.mean(),hz.stddev())
mc_stats = 'MC Mean = %.2f\nMC RMS = %.2f'%(hz_mc.mean(),
hz_mc.stddev())
ax2.set_xlabel(cb_label)
ax2.set_title(title)
ax2.text(0.05,0.95,
'%s\n%s'%(data_stats,mc_stats),
verticalalignment='top',
transform=ax2.transAxes,fontsize=10)
def make_image_plot(self,
subplot,
h,
fig,
fig2,
title,
rpsf68,
rpsf95,
smooth=False,
resid_type=None,
mc_resid=None,
**kwargs):
plt.figure(fig.get_label())
cb_label = 'Counts'
if resid_type == 'significance':
kwargs['vmin'] = -5
kwargs['vmax'] = 5
kwargs['levels'] = [-5.0, -3.0, 3.0, 5.0]
cb_label = 'Significance [$\sigma$]'
elif resid_type == 'fractional':
kwargs['vmin'] = -1.0
kwargs['vmax'] = 1.0
kwargs['levels'] = [-1.0, -0.5, 0.5, 1.0]
cb_label = 'Fractional Residual'
if smooth:
kwargs['beam_size'] = [self._rsmooth, self._rsmooth, 0.0, 4]
axim = h.plot(subplot=subplot, cmap='ds9_b', **kwargs)
h.plot_circle(rpsf68, color='w', lw=1.5)
h.plot_circle(rpsf95, color='w', linestyle='--', lw=1.5)
h.plot_marker(marker='x', color='w', linestyle='--')
ax = h.ax()
ax.set_title(title)
cb = plt.colorbar(axim,
orientation='horizontal',
shrink=0.9,
pad=0.15,
fraction=0.05)
if kwargs.get('zscale', None) is not None:
import matplotlib.ticker
cb.locator = matplotlib.ticker.MaxNLocator(nbins=5)
cb.update_ticks()
cb.set_label(cb_label)
cat = Catalog.get('3fgl')
cat.plot(h,
ax=ax,
src_color='w',
label_threshold=self._srclabels_thresh)
if resid_type is None: return
plt.figure(fig2.get_label())
ax2 = fig2.add_subplot(subplot)
z = h.counts[10:-10, 10:-10]
if resid_type == 'significance':
zproj_axis = Axis.create(-6, 6, 120)
elif resid_type == 'fractional':
zproj_axis = Axis.create(-1.0, 1.0, 120)
else:
zproj_axis = Axis.create(-10, 10, 120)
hz = Histogram(zproj_axis)
hz.fill(np.ravel(z))
nbin = np.prod(z.shape)
hz_mc = Histogram(zproj_axis)
if mc_resid:
for mch in mc_resid:
z = mch.counts[10:-10, 10:-10]
hz_mc.fill(np.ravel(z))
hz_mc /= float(len(mc_resid))
fn = lambda t: 1. / np.sqrt(2 * np.pi) * np.exp(-t**2 / 2.)
hz.plot(label='Data', linestyle='None')
if resid_type == 'significance':
plt.plot(hz.axis().center,
fn(hz.axis().center) * hz.axis().width * nbin,
color='k',
label='Gaussian ($\sigma = 1$)')
hz_mc.plot(label='MC', hist_style='line')
plt.gca().grid(True)
plt.gca().set_yscale('log')
plt.gca().set_ylim(0.5)
ax2.legend(loc='upper right', prop={'size': 10})
data_stats = 'Mean = %.2f\nRMS = %.2f' % (hz.mean(), hz.stddev())
mc_stats = 'MC Mean = %.2f\nMC RMS = %.2f' % (hz_mc.mean(),
hz_mc.stddev())
ax2.set_xlabel(cb_label)
ax2.set_title(title)
ax2.text(0.05,
0.95,
'%s\n%s' % (data_stats, mc_stats),
verticalalignment='top',
transform=ax2.transAxes,
fontsize=10)
sys.exit(0)
nevent = 10000
cm = CompositeModel()
# cm.addModel(KingFn.create(1.0,3.0,0.5*nevent))
cm.addModel(PolyFn.create(3,[0,1,2]))
# cm.addModel(KingFn.create(0.2,3.0,0.5*nevent,3))
print cm.param()
nbin = 80
h0 = Histogram([0,5.0],nbin)
h1 = Histogram([0,5.0],nbin)
h0.fill(cm.rnd(nevent,10.0))
lnlfn = OnOffBinnedLnL(h0._counts,h1._counts,h0._xedges,1.0,cm)
plt.figure()
sig = np.linspace(0.9,1.1,100)
sys.exit(0)
nevent = 10000
cm = CompositeModel()
# cm.addModel(KingFn.create(1.0,3.0,0.5*nevent))
cm.addModel(PolyFn.create(3, [0, 1, 2]))
# cm.addModel(KingFn.create(0.2,3.0,0.5*nevent,3))
print cm.param()
nbin = 80
h0 = Histogram([0, 5.0], nbin)
h1 = Histogram([0, 5.0], nbin)
h0.fill(cm.rnd(nevent, 10.0))
lnlfn = OnOffBinnedLnL(h0._counts, h1._counts, h0._xedges, 1.0, cm)
plt.figure()
sig = np.linspace(0.9, 1.1, 100)
p = cm.param().makeParameterArray(1, sig)
#print lnlfn.eval(p)
#print lnlfn.eval(p[0])
parser.add_argument('--colors', default=None)
parser.add_argument('--ylim_ratio', default='0.6/1.4')
parser.add_argument('--prefix', default='')
args = parser.parse_args()
labels = args.labels.split(',')
if args.colors: colors = args.colors.split(',')
else: colors = ['b', 'g', 'r', 'm', 'c']
ylim_ratio = [float(t) for t in args.ylim_ratio.split('/')]
hist_set = {
'title': 'test',
'ts': Histogram(Axis.create(0., 16., 160)),
'fluxul': Histogram(Axis.create(-12, -8, 100))
}
pwl_hists = []
dm_hists = []
hists = {}
results = []
limits = {}
for f in args.files:
# results = AnalysisResults(f)
