def test_fit_2D(self):
"""
Test 2D Hist.curve_fit() and Hist.spline_fit().
"""
np.random.seed(0)
h = hl.hist(np.random.normal(0, 1, (2, int(1e5))),
bins=50,
range=(-4, 4))
hn = h.normalize()
# curve_fit
def gauss2(x, sigma):
return 1 / (2*np.pi * sigma**2) \
* np.exp (-np.sum (x**2, axis=0) / (2 * sigma**2))
sigma = hn.curve_fit(gauss2)[0][0]
self.assertGreater(.02, np.abs(sigma - 1))
# spline_fit
s = hn.spline_fit()
deltas = np.abs(s(*np.meshgrid(*hn.centers)) - hn.values)
self.assertTrue(
np.all(3e-2 > np.abs(s(*np.meshgrid(*hn.centers)) - hn.values)))
s = hn.spline_fit(log=True)
deltas = np.abs(s(*np.meshgrid(*hn.centers)) - hn.values)
self.assertTrue(
np.all(3e-2 > np.abs(s(*np.meshgrid(*hn.centers)) - hn.values)))
def test_range(self):
"""
Test basic hist() results.
"""
x = np.random.uniform(0, 1, int(1e4))
# default values
h = hl.hist(x)
self.assertFalse(h.log[0])
dx0 = abs(h.bins[0][0] - np.min(x))
dx1 = abs(h.bins[0][-1] - np.max(x))
self.assertTrue(dx0 <= eps)
self.assertTrue(dx1 <= eps)
self.assertEqual(h.n_dim, 1)
self.assertEqual(len(h.n_bins), 1)
self.assertEqual(len(h.bins), 1)
self.assertEqual(h.n_bins[0], len(h.centers[0]))
self.assertEqual(h.n_bins[0], len(h.bins[0]) - 1)
self.assertEqual(
h.n_bins[0],
np.sum((h.bins[0][:-1] < h.centers[0])
& (h.bins[0][1:] > h.centers[0])))
self.assertEqual(h.n_bins[0], len(h.values))
self.assertTrue(
eps >= np.abs(np.sum(h.widths) - (h.range[0][1] - h.range[0][0])))
self.assertTrue(np.max(np.abs(h.widths[0] - h.volumes) < eps))
# log scale
h = hl.hist(x, log=True)
self.assertTrue(h.log[0])
dx0 = abs(h.bins[0][0] - np.min(x))
dx1 = abs(h.bins[0][-1] - np.max(x))
self.assertTrue(dx0 <= eps)
self.assertTrue(dx1 <= eps)
self.assertEqual(
h.n_bins[0],
np.sum((h.bins[0][:-1] < h.centers[0])
& (h.bins[0][1:] > h.centers[0])))
self.assertTrue(
eps >= np.abs(np.sum(h.widths) - (h.range[0][1] - h.range[0][0])))
# simple 3D
x = np.random.uniform(0, 1, (3, int(1e3)))
h = hl.hist(x, range=(0, 1))
self.assertTrue(len(h.bins) == len(h.centers) == h.n_dim)
self.assertEqual(h.values.shape, tuple(h.n_bins))
self.assertTrue(3 * eps >= np.abs(np.sum(h.volumes) - 1))
def aeff(self):
"""
Plot and tabulate Aeff.
"""
import colormaps as cmaps
plt.register_cmap(name='viridis', cmap=cmaps.viridis)
plt.set_cmap(cmaps.viridis)
logEmin, logEmax = 2., 9.
dlogE = 0.1
n_bins_E = (logEmax - logEmin) / dlogE
dcz = 0.01
dOmega = 2 * pi * dcz
n_bins_cz = 2 / dcz
nu = self.nu
nu.cz = -np.sin(nu.trueDec)
w_aeff = 1 / (1e4 * np.log(10)) * nu.ow / nu.trueE / dOmega / dlogE
h_aeff = hl.hist(
(nu.trueE, nu.cz),
w_aeff,
bins=(n_bins_E, n_bins_cz),
range=((10**logEmin, 10**logEmax), (-1, 1)),
log=(True, False),
)
misc.tex_mpl_rc(True)
fig = getfig(aspect=4 / 3., width=5)
ax = fig.add_subplot(111)
fig.subplots_adjust(bottom=.15, left=.15)
result = hl.plot2d(ax,
h_aeff,
cbar=True,
log=True,
vmin=5e-6,
vmax=1e4,
zmin=5e-6)
result['colorbar'].set_label(r'effective area $[\text{m}^2]$')
ax.set_xlabel('neutrino energy [GeV]')
ax.set_ylabel(r'$\cos(\text{zenith})$')
plot_dir = misc.ensure_dir('{0}/plots'.format(self.mode_dir))
savingfig(fig, plot_dir, 'aeff')
bins = h_aeff.bins
filename = '{}/aeff.txt'.format(plot_dir)
prush('-> {} ...'.format(filename))
with open(filename, 'w') as f:
pr = lambda *a, **kw: print(*a, file=f, **kw)
pr('# {:>11}{:>13}{:>16}{:>16}{:>16}'.format(
'E_min[GeV]', 'E_max[GeV]', 'cos(zenith)_min',
'cos(zenith)_max', 'Aeff[m^2]'))
for (Emin, Emax) in izip(bins[0][:-1], bins[0][1:]):
for (czmin, czmax) in izip(bins[1][:-1], bins[1][1:]):
pr('{:13.3e}{:13.3e}{:+16.2f}{:+16.2f}{:16.3e}'.format(
Emin, Emax, czmin, czmax,
h_aeff.get_value(1.001 * Emin, 1e-3 + czmin)))
def test_sum(self):
"""
Test Hist.sum(), Hist.cumsum(), and related methods.
"""
np.random.seed(0)
shape = 4, int(2e4)
h = hl.hist(np.random.uniform(0, 1, shape),
bins=(2, 4, 5, 10),
range=(0, 1))
self.assertEqual(h.sum().values, shape[1])
self.assertEqual(h.sum().volumes, 0)
self.assertEqual(h.project([1, 2]).values.shape, (4, 5))
with self.assertRaises(ValueError):
h.sum([5])
for i in xrange(h.n_dim):
for j in xrange(h.n_dim):
if i == j:
continue
hn = h.normalize([i, j])
self.assertEqual(np.shape(hn.values), np.shape(h.values))
self.assertEqual(np.shape(hn.errors), np.shape(h.errors))
self.assertTrue(
1e-6 > np.max(np.abs(hn.integrate([i, j]).values - 1)))
hn = h.normalize([i, j], integrate=False)
self.assertEqual(np.shape(hn.values), np.shape(h.values))
self.assertEqual(np.shape(hn.errors), np.shape(h.errors))
self.assertTrue(
1e-6 > np.max(np.abs(hn.sum([i, j]).values - 1)))
for i in xrange(h.n_dim):
hn = h.normalize([i], integrate=False)
hncs = hn.cumsum([i])
hncs_last = hncs.get_slice(-1, i)
self.assertTrue(1e-6 > np.max(np.abs(hncs_last.values) - 1))
h = hl.hist(10**np.random.uniform(0, 1, int(1e4)),
bins=4,
range=(1, 10),
log=True)
hn = h.normalize()
hd = h.normalize(density=True)
self.assertGreater(4 * eps, np.abs(hn.integrate().values - 1))
self.assertGreater(4 * eps, np.abs(hd.integrate().values - 1))
self.assertFalse(np.any(hn.values == hd.values))
def test_transpose(self):
"""
Test Hist.T.
"""
h = hl.hist(np.random.uniform(0, 1, (2, int(1e4))),
bins=(4, 10),
range=(0, 1))
hT = h.T
self.assertEqual(h.bins[::-1], hT.bins)
self.assertTrue(np.all(h.values.T == hT.values))
self.assertTrue(np.all(h.errors.T == hT.errors))
def test_rebin(self):
"""
Test Hist.rebin().
"""
# lin bins
h = hl.hist(np.random.uniform(0, 1, int(1e4)), bins=4, range=(0, 1))
hr = h.rebin(0, [0, .5, 1])
self.assertEqual(h.values[:2].sum(), hr.values[0])
self.assertEqual(h.values[2:4].sum(), hr.values[1])
# log bins
h = hl.hist(10**np.random.uniform(0, 2, int(1e4)),
bins=4,
range=(1, 100),
log=True)
hr = h.rebin(0, [1, 10, 100])
self.assertEqual(h.values[:2].sum(), hr.values[0])
self.assertEqual(h.values[2:4].sum(), hr.values[1])
with self.assertRaises(ValueError):
h + hr
def aeff2d (year):
"""
Plot and tabulate Aeff.
"""
import colormaps as cmaps
plt.register_cmap (name='viridis', cmap=cmaps.viridis)
plt.set_cmap (cmaps.viridis)
logEmin, logEmax = 2., 9.
dlogE = 0.1
n_bins_E = (logEmax - logEmin) / dlogE
dcz = 0.01
dOmega = 2 * np.pi * dcz
n_bins_cz = 2 / dcz
#nu = self.nu ..... xn is the mc sample of a certain year. let's start with IC86I and work our way forward.
if year == 86:
xn = data_multi.MC86I()
elif year ==79:
xn = data_multi.MC79()
elif year ==59:
xn = data_multi.MC59()
elif year ==40:
xn = data_multi.MC40()
#Whichever year we have, the following occurs:
nu = arrays.Arrays (dict (
(k,xn[k])
for k in ('ra', 'sinDec', 'sigma', 'logE',
'trueRa', 'trueDec', 'trueE', 'ow')))
#nu = self.nu
nu.cz = -np.sin (nu.trueDec)
w_aeff = 1 / (1e4 * np.log (10)) * nu.ow / nu.trueE / dOmega / dlogE
h_aeff = histlite.hist (
(nu.trueE, nu.cz), w_aeff,
bins=(n_bins_E, n_bins_cz),
range=((10**logEmin, 10**logEmax), (-1, 1)),
log=(True, False),
)
fig = getfig (aspect=4/3., width=5)
ax = fig.add_subplot (111)
fig.subplots_adjust (bottom=.15, left=.15)
result = histlite.plot2d (ax, h_aeff, cbar=True, log=True,
vmin=5e-6, vmax=1e4, zmin=5e-6)
result['colorbar'].set_label (r'effective area $[\textrm{m}^2]$')
ax.set_title('Aeff - IC{}'.format(str(year)))
ax.set_xlabel ('neutrino energy [GeV]')
ax.set_ylabel (r'$\cos(\textrm{zenith})$')
plot_dir = misc.ensure_dir ('/data/user/brelethford/AGN_Core/Plots/data')
fig.savefig(plot_dir+'/aeff{}_2d.pdf'.format(str(year)))
def plotDecBand(bandnum):
fig_band=plt.figure()
ax=plt.gca()
histdec_data=histlite.plot1d(ax,histlite.hist(muex86I_data[bandmask_data[bandnum]],**hargs1), color='r', label='data')
#use gamma=2,bandnum=0
histdec_sim=histlite.plot1d(ax,energyhist_sim(fsim_86_1,2,bandnum), color='b', label='MC')
ax.loglog()
ax.set_xlabel('MuEx')
ax.set_ylabel('events')
ax.legend(loc='upper right')
ax.set_ylim(ymin=0)
ax.set_title('energy pdf in IC86-I, gamma=2: '+decrange[bandnum])
fig_band.savefig(projfolder+'Plots/AGNCore/Stacking/IC86I_energydist: band '+str(bandnum))
def makehist_inj(datafolder):
files = cache.load(datafolder+'gamma2.0.array')
TSs=files['trials']['TS']
chi2fit_flux = fitfun(TSs, df=1., floc=0., fscale=1.)
weib_flux = weibfun(TSs,df=2., floc=0., fscale=1.)
print('background plus injected: median TS = {}'.format(str(np.asscalar(chi2fit_flux.isf(0.5)))))
print ('max TS = {}'.format(str(max(TSs))))
print ('Percentage of TS=0 is: {}'.format(str(1.0-np.float(np.count_nonzero(TSs))/np.float(len(TSs)))))
##Now we make hists of the test statistics ##
flux_hist =histlite.hist(TSs,bins=bins,range=range)
return flux_hist,chi2fit_flux
def test_err(self):
"""
Test error propagation.
"""
h1 = hl.hist([.25, .75], bins=1, range=(0, 1))
h2 = hl.hist([.25, .5, .75], bins=1, range=(0, 1))
self.assertTrue(eps > np.abs(h1.get_errors([.25]) - np.sqrt(2)))
self.assertTrue(1e-6 > np.max(h1.errors - np.sqrt(2)))
self.assertTrue(1e-6 > np.max(h2.errors - np.sqrt(3)))
hadd = h1 + h2
self.assertTrue(1e-6 > np.max(hadd.errors - np.sqrt(2 + 3)))
hadd = 1 + h1
self.assertTrue(1e-6 > np.max(hadd.errors - np.sqrt(2)))
hsub = h2 - h1
self.assertTrue(1e-6 > np.max(hsub.errors - np.sqrt(2 + 3)))
hsub = 1 - h1
self.assertTrue(1e-6 > np.max(hsub.errors - np.sqrt(2)))
hmul = h1 * h2
self.assertTrue(
1e-6 > np.max(hmul.errors -
6 * np.sqrt(1 / np.sqrt(2) + 1 / np.sqrt(3))))
hmul = 2 * h1
self.assertTrue(1e-6 > np.max(hmul.errors - 2 * np.sqrt(2)))
hdiv = h2 / h1
self.assertTrue(
1e-6 > np.max(hdiv.errors -
1.5 * np.sqrt(1 / np.sqrt(2) + 1 / np.sqrt(3))))
hdiv = h1 / 2.
self.assertTrue(1e-6 > np.max(hdiv.errors - 0.5 * np.sqrt(2)))
x = np.random.uniform(0, 1, int(1e4))
w = np.random.uniform(.2, .8, x.size)
hall = hl.hist(x, w, bins=10, range=(0, 1))
hcut = hl.hist(x[w > .5], w[w > .5], bins=10, range=(0, 1))
div = hcut / hall
eff = hcut.efficiency(hall)
self.assertTrue(np.all(eff.errors < div.errors))
def test_sample(self):
"""
Test Hist.sample().
"""
np.random.seed(0)
h = hl.hist(np.random.uniform(0, 1, (2, int(1e3))),
bins=4,
range=(0, 1))
# test sampling in both dimensions
x, y = h.sample(int(1e5))
h2 = hl.hist((x, y), bins=4, range=(0, 1))
deltas = np.abs((h2 / 1e2 - h).values)
self.assertLess(np.max(deltas - h.errors), 0)
self.assertGreater(np.min(x), h.bins[0][0])
self.assertGreater(np.min(y), h.bins[1][0])
self.assertLess(np.max(x), h.bins[0][-1])
self.assertLess(np.max(y), h.bins[1][-1])
# test sampling for a given x
y = h.sample(int(1e5), x[0])
h1 = hl.hist(y, bins=4, range=(0, 1))
deltas = np.abs((h1 / 1e2 / 4).values - h[x[0]].values)
self.assertLess(np.max(deltas - h.errors), 0)
def test_contain(self):
"""
Test Hist.contain() and related methods.
"""
x = np.random.uniform(0, 1, int(3e5))
y = np.random.normal(0, 1, int(3e5))
h = hl.hist((x, y), bins=(4, 101), range=((0, 1), (-4, 4)))
# median should be around 0
hmed = h.median(1)
self.assertGreater(1e-6, np.abs(np.mean(hmed.values)))
# containment hist errors should reflect sigma=1
hcont = h.contain_project(1, n_sigma=1)
self.assertGreater(.05, np.abs(np.mean(hcont.errors) - 1))
def test_other_create(self):
"""
Test special-purpose Hist.hist_like() and Hist.hist_direct().
"""
np.random.seed(0)
x = np.random.uniform(0, 1, (2, int(1e4)))
h = hl.hist(x, bins=10, range=(0, 1))
hd = hl.hist_direct(x, bins=10, range=2 * [(0, 1)])
self.assertTrue(np.all(h.values == hd.values))
self.assertTrue(np.all(h.errors == hd.errors))
h_like_h = hl.hist_like(h, x)
h_like_hd = hl.hist_like(hd, x)
self.assertTrue(np.all(h_like_h.values == h.values))
self.assertTrue(np.all(h_like_hd.values == h.values))
self.assertTrue(np.all(h_like_h.errors == h.errors))
self.assertTrue(np.all(h_like_hd.errors == h.errors))
def makehist(datafolder):
files = [cache.load(datafolder+file) for file in os.listdir(datafolder) if file.endswith('.array')]
TS=[]
for file in files:
for entry in file:
if entry[0]==0:
if entry[1]<0:
TS.append(0)
else:
TS.append(entry[1])
TSs=TS
chi2fit_flux = fitfun(TSs, df=1., floc=0., fscale=1.)
print('background only: median TS = {}'.format(str(np.asscalar(chi2fit_flux.isf(0.5)))))
print ('max TS = {}'.format(str(max(TSs))))
print ('Number of trials is: {}'.format(str(len(TSs))))
print ('Percentage of TS=0 is: {}'.format(str(1.0-np.float(np.count_nonzero(TSs))/np.float(len(TSs)))))
##Now we make hists of the test statistics ##
flux_hist =histlite.hist(TSs,bins=bins,range=range)
return flux_hist,chi2fit_flux
def test_fit_1D(self):
"""
Test 1D Hist.curve_fit() and Hist.spline_fit().
"""
np.random.seed(0)
h = hl.hist(np.random.normal(0, 1, int(1e5)), bins=50, range=(-4, 4))
hn = h.normalize()
# curve_fit
def gauss(x, sigma):
return 1 / (np.sqrt (2*np.pi) * sigma) \
* np.exp (-x**2 / (2 * sigma**2))
sigma = hn.curve_fit(gauss)[0][0]
self.assertGreater(.01, np.abs(sigma - 1))
# spline_fit
s = hn.spline_fit()
self.assertTrue(np.all(1e-2 > np.abs(s(hn.centers) - hn.values)))
s = hn.spline_fit(log=True)
self.assertTrue(np.all(1e-2 > np.abs(s(hn.centers) - hn.values)))
weighteach = ['uniform', 'flux', 'redshift']
TSeach = [TS_uniform, TS_flux, TS_redshift]
chi_each = [chi2fit_uniform, chi2fit_flux, chi2fit_redshift]
weib_each = [weib_uniform, weib_flux, weib_redshift]
for weight, TS, chi2fit in zip(weighteach, TSeach, chi_each):
print('For ' + str(weight) + ', the median TS is: ' +
str(np.asscalar(chi2fit.isf(0.5))))
print(' max TS is: ' + str(max(TS)))
print(' percentage of TS = 0 is: ') + str(
1.0 - np.float(np.count_nonzero(TS)) / np.float(len(TS)))
##Now we make hists of the test statistics ##
bins = 50
range = (0.0, 40.0)
uniform_hist = histlite.hist(TS_uniform, bins=bins, range=range)
flux_hist = histlite.hist(TS_flux, bins=bins, range=range)
redshift_hist = histlite.hist(TS_redshift, bins=bins, range=range)
fig_bckg = plt.figure(figsize=(w, .75 * w))
ax = plt.gca()
colors = ['blue', 'green', 'red']
for chi2_fit, weib_fit, color in zip(chi_each, weib_each, colors):
x = np.linspace(0, 20, 100)
ax.plot(x,
chi2_fit.pdf(x),
linestyle=':',
color=color,
label=r'$\tilde{\chi}^2$') #: df='+str(round(chi2_fit.par[0],2)))
plt.axvline(np.asscalar(chi2_fit.isf(0.5)), color=color)
def energyhist_sim(fsim,gamma,bandnum): weight=simweight(fsim,gamma,mask=bandmask_sim[bandnum]) return histlite.hist (muex86I_sim[bandmask_sim[bandnum]],weights=liveTime86I*weight,**hargs1)
picklefolder = '/data/user/brelethford/Data/SwiftBAT70m/pickle/'
datafolder = '/data/user/brelethford/Output/all_sky_sensitivity/results/single_stacked/'
## Read in background trials previously logged ##
files = [
cache.load(datafolder + file) for file in os.listdir(datafolder)
if file.endswith('.array')
]
bckg_single = []
for file in files:
bckg_single.append(list(file['TS']))
##Now we make hists of the test statistics ##
bins = 80
range = (0.0, 20.0)
single_hist = histlite.Hist.normalize(
histlite.hist(bckg_single[0], bins=bins, range=range))
## Now to plot. ##
fig_bckg = plt.figure(figsize=(w, .75 * w))
ax = plt.gca()
##I'll include a chi squared distribution w/ DOF=1 (and 2, just because). I'll also show the best fitting chi2 dist for each weighting scheme.##
chifit_single = chi2.fit(bckg_single[0])[0]
chi_degs = [1, 2, chifit_single]
colors = ['black', 'gray', 'blue']
for df, color in zip(chi_degs, colors):
x = np.linspace(chi2.ppf(0.01, df), chi2.ppf(0.99999, df), 100)
rv = chi2(df)
chi_dist = rv.pdf(x)
ax.plot(x,
datafolder_flux = '/data/user/brelethford/Output/stacking_sensitivity/2LAC/flux_3yr/background_trials/' bckg_flux = getBckg(datafolder_flux) print (bckg_flux['TS_beta']) #bckg_uniform = cache.load(picklefolder+'bckg_trials_uniform.pickle') #bckg_redshift = cache.load(picklefolder+'bckg_trials_redshift.pickle') #bckg_flux = cache.load(picklefolder+'bckg_trials_flux.pickle') ##Now we make hists of the test statistics ## bins = 80 range = (0.0,20.0) flux_hist =histlite.Hist.normalize(histlite.hist(bckg_flux['TS'],bins=bins,range=range)) ##I'll include a chi squared distribution w/ DOF=1 (and 2, just because). I'll also show the best fitting chi2 dist for each weighting scheme.## chifit_flux = fitfun(bckg_flux['TS'],df=2., floc=0., fscale=1.).par[0] ## Now to plot. ## fig_bckg = plt.figure (figsize=(w, .75*w)) ax=plt.gca() chi_degs = [1,2,chifit_flux] colors=['black','gray','green'] for df,color in zip(chi_degs,colors): x = np.linspace(chi2.ppf(0.01,df),chi2.ppf(0.99999,df), 100) rv = chi2(df) chi_dist=rv.pdf(x)
'TS_beta': TS_beta,
'gamma': np.asarray(gamma)
}
return bckg_trials
datafolder_2LAC = '/data/user/brelethford/Output/stacking_sensitivity/2LAC/flux_3yr/background_trials/'
bckg_2LAC = getBckg(datafolder_2LAC)
print(bckg_2LAC['TS_beta'])
##Now we make hists of the test statistics ##
bins = 100
range = (0.0, 20.0)
h_2LAC = histlite.hist(bckg_2LAC['TS'], bins=bins, range=range)
##I'll include a chi squared distribution w/ DOF=1 (and 2, just because). I'll also show the best fitting chi2 dist for each weighting scheme.##
chi2fit_2LAC = fitfun(bckg_2LAC['TS'], df=2., floc=0., fscale=1.)
#Now for a weibfit on the distribution.
weib_flux_2LAC = weibfun(bckg_2LAC['TS'], df=2., floc=0., fscale=1.)
## Now to plot. ##
fig_bckg = plt.figure(figsize=(w, .75 * w))
ax = plt.gca()
weib_fits = [weib_flux_2LAC]
chi2_fits = [chi2fit_2LAC]
colors = ['blue']
for chi2_fit, weib_fit, color in zip(chi2_fits, weib_fits, colors):
ben_exp = cache.load (datafolder + 'exp.pickle') ben_mc = cache.load (datafolder + 'MC.pickle') bad_exp = cache.load (datafolder + 'exp_bad_sigma.pickle') bad_mc = cache.load (datafolder + 'MC_bad_sigma.pickle') stefan_exp = np.genfromtxt(stefanfolder+'IC86-I_data.txt',names=True) stefan_mc = np.genfromtxt(stefanfolder+'IC86-I_MC.txt',names=True) #extract the sigmas ben_exp_sigma =ben_exp['sigma'] ben_mc_sigma =ben_mc['sigma'] ben_hist_exp = histlite.hist(ben_exp_sigma,bins=100,log=True) ben_hist_mc = histlite.hist(ben_mc_sigma,bins=100,log=True) bad_exp_sigma =bad_exp['sigma'] bad_mc_sigma =bad_mc['sigma'] bad_hist_exp = histlite.hist(bad_exp_sigma,bins=100,log=True) bad_hist_mc = histlite.hist(bad_mc_sigma,bins=100,log=True) stefan_exp_sigma = stefan_exp['sigma'] stefan_mc_sigma = stefan_mc['sigma'] stefan_hist_exp = histlite.hist(stefan_exp_sigma,bins=100,log=True) stefan_hist_mc = histlite.hist(stefan_mc_sigma,bins=100,log=True) ## Now to plot. ##
datafolder_flux = '/data/user/brelethford/Output/stacking_sensitivity/2LAC/flux/background_trials/'
bckg_flux = getBckg(datafolder_flux)
print(bckg_flux['TS_beta'])
#bckg_uniform = cache.load(picklefolder+'bckg_trials_uniform.pickle')
#bckg_redshift = cache.load(picklefolder+'bckg_trials_redshift.pickle')
#bckg_flux = cache.load(picklefolder+'bckg_trials_flux.pickle')
##Now we make hists of the test statistics ##
bins = 80
range = (0.0, 20.0)
flux_hist = histlite.Hist.normalize(
histlite.hist(bckg_flux['TS'], bins=bins, range=range))
##I'll include a chi squared distribution w/ DOF=1 (and 2, just because). I'll also show the best fitting chi2 dist for each weighting scheme.##
chifit_flux = fitfun(bckg_flux['TS'], df=2., floc=0., fscale=1.).par[0]
## Now to plot. ##
fig_bckg = plt.figure(figsize=(w, .75 * w))
ax = plt.gca()
chi_degs = [1, 2, chifit_flux]
colors = ['black', 'gray', 'green']
for df, color in zip(chi_degs, colors):
x = np.linspace(chi2.ppf(0.01, df), chi2.ppf(0.99999, df), 100)
rv = chi2(df)
chi_dist = rv.pdf(x)
ax.plot(x,
true86I = mc86I['trueE']
weight_86 = mc86I['ow'] * true86I**(-2)
weight_79 = mc79['ow'] * true79**(-2)
weight_59 = mc59['ow'] * true59**(-2)
weight_40 = mc40['ow'] * true40**(-2)
bins = 80
bins2d = (80, 80)
### Energy ###
energyrange = (10, 1e7)
fig_energyhist = plt.figure(figsize=(w, .75 * w))
ax = plt.gca()
h_energy_86 = histlite.hist(muex86I_data,
bins=bins,
range=energyrange,
log=True)
h_energy_79 = histlite.hist(muex79_data,
bins=bins,
range=energyrange,
log=True)
h_energy_59 = histlite.hist(mue59_data, bins=bins, range=energyrange, log=True)
h_energy_40 = histlite.hist(mue40_data, bins=bins, range=energyrange, log=True)
histlite.plot1d(ax, h_energy_40, histtype='step', label='IC40', color='purple')
histlite.plot1d(ax, h_energy_59, histtype='step', label='IC59', color='green')
histlite.plot1d(ax, h_energy_79, histtype='step', label='IC79', color='blue')
histlite.plot1d(ax, h_energy_86, histtype='step', label='IC86', color='red')
#ax.set_title("4yr PS data - energy")
ax.set_xlabel(r"Energy Proxy (GeV)")
ax.set_ylabel("Counts per bin")
#ax.set_xlabel(r'flux $[10^{-12} ergs/s/cm^2]$')
#plt.subplots_adjust (left=.2, bottom=.2)
#ax.set_ylabel(r'z')
#plt.legend(loc='upper right', prop=propsmall)
#fig_redshiftxflux.savefig('/data/user/brelethford/AGN_Core/Plots/flux_vs_z.pdf')
#
#print ('Number of sources = ' + str(len(src_ra)))
#It's also probably worth having a sindec hist of the sources, but I want to weight the sources to each of the three schemes. First let's get in sindec...
sindec = np.sin(src_dec)
#Now I have to establish three hists - one for each weighting scheme...
bins = 100
range = (-1.0, 1.0)
unif_hist = histlite.Hist.normalize(histlite.hist(sindec,
bins=bins,
range=range),
integrate=True)
flux_hist = histlite.Hist.normalize(histlite.hist(sindec,
weights=np.array(flux),
bins=bins,
range=range),
integrate=True)
redshift_hist = histlite.Hist.normalize(histlite.hist(sindec,
weights=np.power(
redshift, -2),
bins=bins,
range=range),
integrate=True)
#Now plot.
nch59_sim, nch59_data = cache.load(filename_pickle + "IC59/NCh.pickle")
ra40_sim, sindec40_sim, ra40_data, sindec40_data, ra40_true, sindec40_true, energy40_true, mue40_sim, mue40_data, sigma40_sim, sigma40_data, OneWeight_IC40, dpsi_IC40 = cache.load(
filename_pickle + "IC40/coords.pickle")
nch40_sim, nch40_data = cache.load(filename_pickle + "IC59/NCh.pickle")
bins = 80
### Energy ###
fig_energyhist = plt.figure(figsize=(w, .75 * w))
ax = plt.gca()
h_energy_86 = histlite.Hist.normalize(
histlite.hist(muex86I_data, bins=bins, log=True))
h_energy_79 = histlite.Hist.normalize(
histlite.hist(muex79_data, bins=bins, log=True))
h_energy_59 = histlite.Hist.normalize(
histlite.hist(mue59_data, bins=bins, log=True))
h_energy_40 = histlite.Hist.normalize(
histlite.hist(mue40_data, bins=bins, log=True))
histlite.plot1d(ax, h_energy_79, histtype='step', label='IC79', color='blue')
histlite.plot1d(ax, h_energy_86, histtype='step', label='IC86', color='red')
histlite.plot1d(ax, h_energy_59, histtype='step', label='IC59', color='green')
histlite.plot1d(ax, h_energy_40, histtype='step', label='IC40', color='purple')
ax.set_title("4yr PS data - energy")
ax.set_xlabel(r"logE (GeV)")
ax.set_ylabel("Relative Abundance")
ax.loglog()
weight_3yr = mc3yr['ow'] * true3yr**(-2)
weight_86 = mc86I['ow'] * true86I**(-2)
weight_79 = mc79['ow'] * true79**(-2)
weight_59 = mc59['ow'] * true59**(-2)
weight_40 = mc40['ow'] * true40**(-2)
bins = 80
bins2d = (80, 80)
### Energy ###
energyrange = (10, 1e7)
fig_energyhist = plt.figure(figsize=(w, .75 * w))
ax = plt.gca()
h_energy_3yr = histlite.hist(muex3yr_data,
bins=bins,
range=energyrange,
log=True)
h_energy_86 = histlite.hist(muex86I_data,
bins=bins,
range=energyrange,
log=True)
h_energy_79 = histlite.hist(muex79_data,
bins=bins,
range=energyrange,
log=True)
h_energy_59 = histlite.hist(mue59_data, bins=bins, range=energyrange, log=True)
h_energy_40 = histlite.hist(mue40_data, bins=bins, range=energyrange, log=True)
histlite.plot1d(ax, h_energy_40, histtype='step', label='IC40', color='blue')
histlite.plot1d(ax, h_energy_59, histtype='step', label='IC59', color='green')
histlite.plot1d(ax, h_energy_79, histtype='step', label='IC79', color='red')
histlite.plot1d(ax, h_energy_86, histtype='step', label='IC86I', color='cyan')
sindec_data = exp['sinDec']
sindec_sim = mc['sinDec']
sigma_data = exp['sigma']
sigma_sim = mc['sigma']
weight = mc['ow'] * true**(-2)
bins = 80
bins2d = (80, 80)
### Energy ###
energyrange = (1e3, 200000)
fig_energyhist = plt.figure(figsize=(w, .75 * w))
ax = plt.gca()
h_energy = histlite.hist(muex_data, bins=bins, range=energyrange, log=True)
histlite.plot1d(ax, h_energy, histtype='step', label='MESE', color='k')
#ax.set_title("4yr PS data - energy")
ax.set_xlabel(r"Energy Proxy (GeV)")
ax.set_ylabel("Counts per bin")
ax.loglog()
ax.set_ylim(1, 1e2)
plt.subplots_adjust(left=.2, bottom=.2)
ax.legend(loc="best", prop=propsmall)
fig_energyhist.savefig(plotfolder + 'energy_hists_MESE.png')
###Declination###
decrange = (-1., 1.)
fig_dechist = plt.figure(figsize=(w, .75 * w))
ax = plt.gca()
nch40_sim, nch40_data = cache.load(filename_pickle + "IC40/NCh.pickle")
bigrange = (-1., 1.)
# (np.sin(np.radians(160.)),np.sin(np.radians(240.)))
bins = 20
### zen true and zen reco for dpsi > 160 deg ###
###Declination###
fig_dechist = plt.figure(figsize=(w, .75 * w))
ax = plt.gca()
h_dec_86 = histlite.Hist.normalize(
histlite.hist(sindec86I_sim[(np.degrees(dpsi_IC86I) > 160.)],
bins=bins,
range=bigrange))
h_dec_79 = histlite.Hist.normalize(
histlite.hist(sindec79_sim[(np.degrees(dpsi_IC79) > 160.)],
bins=bins,
range=bigrange))
h_dec_59 = histlite.Hist.normalize(
histlite.hist(sindec59_sim[(np.degrees(dpsi_IC59) > 160.)],
bins=bins,
range=bigrange))
h_dec_40 = histlite.Hist.normalize(
histlite.hist(sindec40_sim[(np.degrees(dpsi_IC40) > 160.)],
bins=bins,
range=bigrange))
histlite.plot1d(ax, h_dec_40, histtype='step', label='IC40', color='purple')
bckg_flux = getBckg(datafolder_flux) bckg_redshift = getBckg(datafolder_redshift) print (bckg_uniform['TS_beta']) print (bckg_flux['TS_beta']) print (bckg_redshift['TS_beta']) #bckg_uniform = cache.load(picklefolder+'bckg_trials_uniform.pickle') #bckg_redshift = cache.load(picklefolder+'bckg_trials_redshift.pickle') #bckg_flux = cache.load(picklefolder+'bckg_trials_flux.pickle') ##Now we make hists of the test statistics ## bins = 1000 range = (0.0,20.0) uniform_hist =histlite.hist(bckg_uniform['TS'],bins=bins,range=range) redshift_hist =histlite.hist(bckg_redshift['TS'],bins=bins,range=range) flux_hist =histlite.hist(bckg_flux['TS'],bins=bins,range=range) #I'll include a chi squared distribution w/ DOF=1 (and 2, just because). I'll also show the best fitting chi2 dist for each weighting scheme.## chi2fit_uniform = fitfun(bckg_uniform['TS'],df=2., floc=0., fscale=1.) chi2fit_redshift = fitfun(bckg_redshift['TS'],df=2., floc=0., fscale=1.) chi2fit_flux = fitfun(bckg_flux['TS'],df=2., floc=0., fscale=1.) ## Now to plot. ## fig_bckg = plt.figure (figsize=(w, .75*w)) ax=plt.gca() chi2_fits = [chi2fit_uniform,chi2fit_redshift,chi2fit_flux] colors=['blue','red','green'] for chi2_fit,color in zip(chi2_fits,colors):
def plotpull2d(year):
#The following currently devoted to making energy and angular error plots for a year of data.
# init likelihood class
mc = np.load(filename_pickle + "IC{}_MC.npy".format(year))
dpsi = [
astro.angular_distance(mc['trueRa'][i], mc['trueDec'][i], mc['ra'][i],
np.arcsin(mc['sinDec'][i]))
for i in range(len(mc))
]
mc_corr = np.load(filename_pickle + "IC{}_corrected_MC.npy".format(year))
dpsi_corr = [
astro.angular_distance(mc_corr['trueRa'][i],
mc_corr['trueDec'][i], mc_corr['ra'][i],
np.arcsin(mc_corr['sinDec'][i]))
for i in range(len(mc_corr))
]
colors = ['g'] #['b','g','y','r']
colors_corr = ['r'] #['b','g','y','r']
gamma = np.array([2.]) #np.linspace(1., 2.7, 4)
fig_pull, (ax) = plt.subplots(ncols=1, figsize=(10, 5))
dec = np.arcsin(mc["sinDec"])
angdist = np.degrees(dpsi)
dec_corr = np.arcsin(mc_corr["sinDec"])
angdist_corr = np.degrees(dpsi_corr)
#medians = [misc.weighted_median(np.log10(np.degrees(mc["sigma"])/(angdist)),mc["ow"] * mc["trueE"]**(-g)) for g in gamma]
#for g in range(len(gamma)):
# ax1.axvline(medians[g], color=colors[g], linestyle = 'dotted', alpha = 0.3)
#ax1.hist([np.log10(np.degrees(mc_corr["sigma"])/(angdist_corr)) for i in gamma], label = [r"'Corrected' $\Delta \psi / \sigma$ - $\gamma={0:.1f}$".format(g) for g in gamma], linestyle = 'dotted',
# weights=[mc_corr["ow"] * mc_corr["trueE"]**(-g) for g in gamma], color=[colors_corr[g] for g in range(len(gamma))],
# histtype="step", bins=100, normed=True)
#medians = [misc.weighted_median(np.log10(np.degrees(mc_corr["sigma"])/(angdist_corr)),mc["ow"] * mc["trueE"]**(-g)) for g in gamma]
#for g in range(len(gamma)):
# ax1.axvline(medians[g], color=colors_corr[g], marker='+',alpha = 0.3)
#ax1.set_title(r"Reco MC $\Delta \psi / \sigma$ Check - IC{}".format(str(year)))
#ax1.set_xlabel(r"log$\Delta \psi / \sigma_{ang}$")
#ax1.set_ylabel("Relative Abundance")
#ax1.set_ylim(0,1.5)
#ax1.set_xlim(-2.5,2.5)
#ax1.axvline(x=0, color='k')
#ax1.axvline(x=np.log10(1./1.1774), color='k')
#set gamma
g = 2.0
pull = np.log10(np.degrees(mc["sigma"]) / (angdist))
#Need to calc the median of pull for each energyrange:
pullrange = (-4, 4)
erange = (3, 8)
ebins = 50
pullbins = 200
eedges = np.linspace(erange[0], erange[1], ebins + 1)
ezones = zip(eedges[:-1], eedges[1:])
emid = [np.median(ezone) for ezone in ezones]
emasks = [
(np.log10(mc["trueE"]) > ezone[0]) & (np.log10(mc["trueE"]) < ezone[1])
for ezone in ezones
]
medians = [
misc.weighted_median(pull[emask],
mc["ow"][emask] * mc["trueE"][emask]**(-g))
for emask in emasks
]
#pull_corr = [(np.degrees(mc_corr["sigma"]))/(angdist_corr) for i in gamma]
hpull = histlite.hist((np.log10(mc['trueE']), pull),
weights=mc["ow"] * mc["trueE"]**(-g),
bins=(ebins, pullbins),
range=(erange, pullrange),
log=(False, False))
histlite.plot2d(
ax,
hpull.normalize([-1]),
label=[r"$\Delta \psi / \sigma$ - $\gamma={0:.1f}$".format(g)],
cbar=True,
cmap='jet',
color=colors[0])
ax.scatter(emid, medians, color='white')
ax.axhline(y=np.log10(1.1774), color='white', linestyle='dashed')
#ax.hist([(np.degrees(mc_corr["sigma"]))/(angdist_corr) for i in gamma], label = [r"'Corrected' $\Delta \psi / \sigma$ - $\gamma={0:.1f}$".format(g) for g in gamma], linestyle = 'dotted',
# weights=[mc_corr["ow"] * mc_corr["trueE"]**(-g) for g in gamma], color=[colors_corr[g] for g in range(len(gamma))],
# histtype="step", bins=1000, range = (0,5), normed=True)
ax.set_title("Pull for IC{}".format(str(year)))
ax.legend(loc="upper right")
ax.set_xlim(3, 8)
ax.set_xlabel("log(trueE[GeV])")
ax.set_ylim(-2, 2)
ax.set_ylabel(r"log($\Delta \psi / \sigma_{ang}$)")
fig_pull.savefig(filename_plots + 'opp_pull_IC{}.pdf'.format(str(year)))
