def plot(self, cmap=pylab.cm.hot, title=''):
my_matrix = np.copy(self.matrix)
for i in range(self.namps):
my_matrix[i][i] = 0
win = plot.Window()
fig, axes = win.fig, win.axes[-1]
foo = axes.imshow(my_matrix, interpolation='nearest', cmap=cmap)
pylab.xlabel('victim')
pylab.ylabel('aggressor')
axes.set_xticks(range(self.namps))
axes.set_yticks(range(self.namps))
axes.set_xticklabels(['%i' % i for i in range(1, self.namps + 1)])
axes.set_yticklabels(['%i' % i for i in range(1, self.namps + 1)])
cbar = fig.colorbar(foo)
axes.set_title(title)
#
# Unbinned mode
#
interval0 = num.sort(ra.random(20))
interval1 = num.sort(ra.random(100)) + interval0[-1]
interval2 = num.sort(ra.random(20)) + interval1[-1]
seq = num.concatenate((interval0, interval1, interval2))
bb = BayesianBlocks.BayesianBlocks(seq)
bbp = BayesianBlocks_python.BayesianBlocks(seq)
xr = (0, 3)
bins = 50
binsize = float(xr[1] - xr[0])/bins
win0 = plot.Window(0)
plot.histogram(seq, xrange=xr, bins=bins)
for ncpPrior in range(1, 10):
xx, yy = bb.lightCurve(ncpPrior)
plot.curve(xx, num.array(yy)*binsize, color='r', linewidth=3)
xxp, yyp = bbp.lightCurve(ncpPrior)
plot.curve(xxp, num.array(yyp)*binsize, color='b', linewidth=1)
#
# Exposure weighting in unbinned mode
#
exposures = ra.random(len(seq))
bb.setCellSizes(exposures)
def plot(self,
xrange=None,
interactive=False,
bins=100,
win=None,
subplot=(1, 1, 1),
figsize=None,
add_labels=False,
frameLabels=False,
amp=1,
title=''):
pylab_interactive_state = pylab.isinteractive()
pylab.interactive(interactive)
if win is None:
if frameLabels:
xlabel = 'Bias Corrected Event Signal (DN)'
ylabel = 'Entries / bin'
else:
xlabel, ylabel = None, None
win = plot.Window(subplot=subplot,
figsize=figsize,
xlabel=xlabel,
ylabel=ylabel,
size='large')
else:
win.select_subplot(*subplot)
if frameLabels:
bbox = win.axes[-1].get_position()
points = bbox.get_points()
points[0] += 0.025
points[1] += 0.025
bbox.set_points(points)
win.axes[-1].set_position(bbox)
if xrange is not None:
self.xrange = xrange
logscale = True
if max(self.signals) <= 0:
logscale = False
try:
hist = pylab.hist(self.signals,
bins=bins,
range=self.xrange,
histtype='bar',
color='b',
log=logscale)
yrange = 1, max(hist[0]) * 1.5
plot.setAxis(self.xrange, yrange)
except:
return win
if add_labels:
pylab.xlabel('Bias Corrected Event Signal (DN)')
pylab.ylabel('Entries / bin')
x = (hist[1][1:] + hist[1][:-1]) / 2.
xx = np.linspace(x[0], x[-1], 1000)
pylab.plot(xx,
fe55_lines(xx, *self.pars),
'r--',
markersize=3,
linewidth=1)
pylab.annotate(("Amp %i\nGain=%.2f e-/DN") % (amp, self.gain),
(0.475, 0.8),
xycoords='axes fraction',
size='x-small')
pylab.interactive(pylab_interactive_state)
return win
def fe55_gain_fitter(signals,
ccdtemp=-95,
make_plot=False,
xrange=None,
bins=100,
hist_nsig=10,
title='',
plot_filename=None,
interactive=True,
ylog=True):
"""
Function to fit the distribution of charge cluster DN values from
a Fe55 dataset. A two Gaussian model of Mn K-alpha and K-beta
lines is assumed with the ratio between the K-alpha and K-beta
energies fixed at 5.889/6.49 and the the Gaussian width of the
lines set equal.
The gain (Ne/DN), location and sigma of the K-alpha peak (in units
of DN) are returned as a tuple.
If make_plot=True, then a matplotlib plot of the distribution and
fit is displayed.
If xrange is not None, then that 2-element tuple is used as the
histogram x-range.
If xrange is None, then the histogram x-range is set to
+/- hist_nsig*clipped_stdev about the median of the signal
distribution.
"""
flags = afwMath.MEDIAN | afwMath.STDEVCLIP
try:
stats = afwMath.makeStatistics(signals.tolist(), flags)
except:
print signals
raise
median = stats.getValue(afwMath.MEDIAN)
stdev = stats.getValue(afwMath.STDEVCLIP)
if xrange is None:
# Set range of histogram to include both Kalpha and Kbeta peaks.
xmin = max(median - hist_nsig * stdev, 200)
xmax = min(median * 1785. / 1620. + hist_nsig * stdev, 1000)
xrange = xmin, xmax
# Save pylab interactive state.
pylab_interactive_state = pylab.isinteractive()
# Determine distribution mode and take that as the location of the
# Kalpha peak
hist = np.histogram(signals, bins=bins, range=xrange)
xpeak = hist[1][np.where(hist[0] == max(hist[0]))][0]
xrange = max(0, xpeak - 200), xpeak * 1785. / 1620. + 200
hist = np.histogram(signals, bins=bins, range=xrange)
yrange = 1, max(hist[0]) * 1.5
if make_plot:
if interactive:
pylab.ion()
else:
pylab.ioff()
# fig = pylab.figure()
# axes = fig.add_subplot(111)
win = plot.Window()
hist = pylab.hist(signals,
bins=bins,
range=xrange,
histtype='bar',
color='b',
log=ylog)
if ylog:
plot.setAxis(xrange, yrange)
else:
pylab.ioff()
# hist = np.histogram(signals, bins=bins, range=xrange)
x = (hist[1][1:] + hist[1][:-1]) / 2.
y = hist[0]
ntot = sum(y)
#
# Starting values for two Gaussian fit. The relative
# normalizations are initially set at the expected line ratio
# of K-alpha/K-beta = 0.88/0.12. The relative peak locations
# and relative widths are fixed in fe55_lines(...) above.
#
p0 = (ntot * 0.88, median, stdev / 2., ntot * 0.12)
pars, _ = scipy.optimize.curve_fit(fe55_lines, x, y, p0=p0)
kalpha_peak, kalpha_sigma = pars[1], pars[2]
fe55_yield = Fe55Yield(ccdtemp)
gain = fe55_yield.alpha()[0] / kalpha_peak
if make_plot:
pylab.xlabel('Bias Corrected Event Signal (DN)')
pylab.ylabel('Entries / bin')
xx = np.linspace(x[0], x[-1], 1000)
pylab.plot(xx, fe55_lines(xx, *pars), 'r--', markersize=3, linewidth=1)
pylab.annotate(("K-alpha peak = %i DN\n\n" + "Gain = %.2f e-/DN\n\n") %
(kalpha_peak, gain), (0.5, 0.7),
xycoords='axes fraction')
win.set_title(title)
if plot_filename is not None:
pylab.savefig(plot_filename)
# Restore pylab interactive state.
pylab.interactive(pylab_interactive_state)
return gain, kalpha_peak, kalpha_sigma