def stats(self, alpha=0.05, start=0, batches=100):
"""
Generate posterior statistics for node.
"""
from utils import hpd, quantiles
from numpy import sqrt
try:
trace = np.squeeze(np.array(self.trace(), float)[start:])
n = len(trace)
if not n:
print 'Cannot generate statistics for zero-length trace in', self.__name__
return
return {
'n': n,
'standard deviation': trace.std(0),
'mean': trace.mean(0),
'%s%s HPD interval' % (int(100*(1-alpha)),'%'): hpd(trace, alpha),
'mc error': batchsd(trace, batches),
'quantiles': quantiles(trace)
}
except:
print 'Could not generate output statistics for', self.__name__
return
def histosAndQuantiles(tfile = None, dir = "") :
histos = {}
quantiles = {}
for tkey in tfile.Get(dir).GetListOfKeys() :
key = tkey.GetName()
histos[key] = tfile.Get("/%s/%s"%(dir, key))
quantiles[key] = utils.quantiles(histos[key], sigmaList = [-1.0, 0.0, 1.0])
return histos,quantiles
def histosAndQuantiles(tfile=None, dir=""):
histos = {}
quantiles = {}
for tkey in tfile.Get(dir).GetListOfKeys():
key = tkey.GetName()
histos[key] = tfile.Get("/%s/%s" % (dir, key))
quantiles[key] = utils.quantiles(histos[key],
sigmaList=[-1.0, 0.0, 1.0])
return histos, quantiles
def stats(self, alpha=0.05, start=0, batches=100, chain=None):
"""
Generate posterior statistics for node.
:Parameters:
alpha : float
The alpha level for generating posterior intervals. Defaults to
0.05.
start : int
The starting index from which to summarize (each) chain. Defaults
to zero.
batches : int
Batch size for calculating standard deviation for non-independent
samples. Defaults to 100.
chain : int
The index for which chain to summarize. Defaults to None (all
chains).
"""
from utils import hpd, quantiles
from numpy import sqrt
try:
trace = np.squeeze(np.array(self.trace(burn=start, chain=chain), float))
n = len(trace)
if not n:
print 'Cannot generate statistics for zero-length trace in', self.__name__
return
return {
'n': n,
'standard deviation': trace.std(0),
'mean': trace.mean(0),
'%s%s HPD interval' % (int(100*(1-alpha)),'%'): hpd(trace, alpha),
'mc error': batchsd(trace, batches),
'quantiles': quantiles(trace)
}
except:
print 'Could not generate output statistics for', self.__name__
return
def histoLines(args = {}, key = None, histo = None) :
hLine = r.TLine(); hLine.SetLineColor(args["quantileColor"])
bestLine = r.TLine(); bestLine.SetLineColor(args["bestColor"])
errorLine = r.TLine(); errorLine.SetLineColor(args["errorColor"])
q = utils.quantiles(histo, sigmaList = [-1.0, 0.0, 1.0])
min = histo.GetMinimum()
max = histo.GetMaximum()
best = args["bestDict"][key]
error = args["errorDict"][key] if "errorDict" in args else None
out = []
out.append(hLine.DrawLine(q[1], min, q[1], max))
out.append(hLine.DrawLine(q[0], min, q[0], max))
out.append(hLine.DrawLine(q[2], min, q[2], max))
out.append(bestLine.DrawLine(best, min, best, max))
if error!=None : out.append(errorLine.DrawLine(best - error, max/2.0, best + error, max/2.0))
if "print" in args and args["print"] : print "%20s: %g + %g - %g"%(histo.GetName(), best, q[2]-best, best-q[0])
return out
def createHistos(var,filename,isData,histos,q):
#Getting histograms labeling
rec = var+'_rec'
wgt = rec+'_wgt'
gen = var+'_gen'
mig = var+'_migration'
#open file
fIn=ROOT.TFile.Open(filename)
#loop over events in the tree and fill histos
tree=fIn.Get('DileptonInfo')
for i in xrange(0,tree.GetEntriesFast()):
tree.GetEntry(i)
#select only emu events
if tree.EvCat != -11*13 : continue
if not isData:
if tree.GenLpPt == 0 or tree.GenLmPt == 0: continue
#base weight: BR fix for ttbar x pileup x lepton selection x xsec weight
baseWeight = tree.Weight[0]*tree.Weight[1]*tree.Weight[4] #*tree.XSWeight
#event weight
weight = 1 if isData else baseWeight
#positive lepton
lp=ROOT.TLorentzVector()
lp.SetPtEtaPhiM(tree.LpPt,tree.LpEta,tree.LpPhi,0.)
glp=ROOT.TLorentzVector()
glp.SetPtEtaPhiM(tree.GenLpPt,tree.GenLpEta,tree.GenLpPhi,0.)
#negative lepton
lm=ROOT.TLorentzVector()
lm.SetPtEtaPhiM(tree.LmPt,tree.LmEta,tree.LmPhi,0.)
glm=ROOT.TLorentzVector()
glm.SetPtEtaPhiM(tree.GenLmPt,tree.GenLmEta,tree.GenLmPhi,0.)
#charged lepton pair - pt
ll=ROOT.TLorentzVector()
ll = lp + lm
gll=ROOT.TLorentzVector()
gll = glp + glm
#fill the histograms according to the distrubution variable
#pT positive lepton
if var == 'ptpos':
histos[rec].Fill(lp.Pt(),weight)
binWidth = histos[wgt].GetXaxis().GetBinWidth(histos[wgt].GetXaxis().FindBin(lp.Pt() ) )
histos[wgt].Fill(lp.Pt(),weight/binWidth)
if not isData:
histos[gen].Fill(glp.Pt(),weight)
histos[mig].Fill(glp.Pt(),lp.Pt(),weight)
#Second distribution: Pt(l+l-) = ll.Pt
if var == 'ptll':
histos[rec].Fill(ll.Pt(),weight)
binWidth = histos[wgt].GetXaxis().GetBinWidth(histos[wgt].GetXaxis().FindBin(ll.Pt() ) )
histos[wgt].Fill(ll.Pt(),weight/binWidth)
if not isData:
histos[gen].Fill(gll.Pt(),weight)
histos[mig].Fill(gll.Pt(),ll.Pt(),weight)
#Third distribution: M(l+l-) = ll.M
if var == 'mll':
histos[rec].Fill(ll.M(),weight)
binWidth = histos[wgt].GetXaxis().GetBinWidth(histos[wgt].GetXaxis().FindBin(ll.M() ) )
histos[wgt].Fill(ll.M(),weight/binWidth)
if not isData:
histos[gen].Fill(gll.M(),weight)
histos[mig].Fill(gll.M(),ll.M(),weight)
#Fourth distribution: E(l+)+E(l-) = lp.E() + lm.E()
if var == 'EposEm':
histos[rec].Fill(lp.E() + lm.E(),weight)
binWidth = histos[wgt].GetXaxis().GetBinWidth(histos[wgt].GetXaxis().FindBin(lp.E() + lm.E() ) )
histos[wgt].Fill(lp.E() + lm.E(),weight/binWidth)
if not isData:
histos[gen].Fill(glp.E() + glm.E(),weight)
histos[mig].Fill(glp.E() + glm.E(),lp.E() + lm.E(),weight)
#Fifth distribution: Pt(l+)+Pt(l-) = lp.Pt() + lm.Pt()
if var == 'ptposptm':
histos[rec].Fill(lp.Pt() + lm.Pt(),weight)
binWidth = histos[wgt].GetXaxis().GetBinWidth(histos[wgt].GetXaxis().FindBin(lp.Pt() + lm.Pt() ) )
histos[wgt].Fill(lp.Pt() + lm.Pt(),weight/binWidth)
if not isData:
histos[gen].Fill(glp.Pt() + glm.Pt(),weight)
histos[mig].Fill(glp.Pt() + glm.Pt(),lp.Pt() + lm.Pt(),weight)
#close file
fIn.Close()
# Gets quantiles from histos if q is True
# GetQuantiles just works for TH1 not TH2, therefore h != migration
if q == True:
#print 'Getting quantiles from %s' %filename
for h in histos:
if h == gen:
q_gen=[]
q_gen=utils.quantiles(histos[h])
#print 'quantiles lenght for gen %s' %len(q_gen)
#for i in xrange(0,len(q_gen)): print q_gen[i]
#there should be an easier way to rebin the histograms just using the function rebin that I defined in utils.py but it doesn't seem to be working
#histos[h]=utils.rebin(hist,q_gen[h])
#if h == wgt or h == rec:
if h == rec:
q_rec=[]
q_rec=utils.quantiles(histos[h])
for i in xrange(0,len(q_rec)): print q_rec[i]
return q_gen,q_rec
def plot_reads(Reads, xts, motiflen=None, q=5, quantile=True, bounds=None, subtitle=None, cellnames=None, title=None, hist=False):
"""
reads is a list of list of binary tuples of lists of numpy array with read counts. The outermost
list is over different cell types. The next inner list
is over different TSS dist thresholds. The binary tuple is for same / opposite strands.
"""
width = Reads[0][0][0][0].size
figure = plot.figure()
xvals = np.arange(-width/2,width/2)
numcols = len(Reads)
numrows = len(Reads[0])
if hist:
numrows = 2*numrows
colors = colorwheel(q)
for cellidx, reads in enumerate(Reads):
for index, (xt,read) in enumerate(zip(xts,reads)):
if quantile:
quantized = utils.quantiles(xt, q=q)
else:
quantized = utils.quantize(xt, q=q, bounds=bounds)
same = [np.mean([read[0][idx] for idx in quant if read[0][idx].size==width],0) for quant in quantized]
opp = [-1*np.mean([read[1][idx] for idx in quant if read[1][idx].size==width],0) for quant in quantized]
if hist:
subplot = figure.add_subplot(numrows,numcols,2*index*numcols+cellidx+1)
else:
subplot = figure.add_subplot(numrows,numcols,index*numcols+cellidx+1)
subplot = remove_spines(subplot)
fwd = [subplot.plot(xvals, s, color=c, linestyle='-', linewidth=0.5) for s,c in zip(same,colors)]
rev = [subplot.plot(xvals, o, color=c, linestyle='-', linewidth=0.5) for o,c in zip(opp,colors)]
subplot.axhline(0, linestyle='--', linewidth=0.2)
subplot.axvline(0, linestyle='--', linewidth=0.2)
if motiflen:
subplot.axvline(motiflen-1, linestyle='--', c='g', linewidth=0.2)
xmin = xvals[0]
xmax = xvals[-1]
ymax = max([s.max() for s in same])
ymin = min([o.min() for o in opp])
subplot.axis([xmin, xmax, ymin, ymax])
for text in subplot.get_xticklabels():
text.set_fontsize(7)
text.set_verticalalignment('center')
ytick_locs = list(np.linspace(np.round(ymin,2),np.round(ymax,2),5))
if 0 not in ytick_locs:
ytick_locs.append(0)
ytick_locs.sort()
ytick_labels = tuple(['%.2f'%s for s in ytick_locs])
subplot.set_yticks(ytick_locs)
subplot.set_yticklabels(ytick_labels, color='k', fontsize=6, horizontalalignment='right')
if subtitle and cellidx==0:
bbox = subplot.get_position()
xloc = bbox.xmin/3.
yloc = (bbox.ymax+bbox.ymin)/2.
plot.text(xloc, yloc, subtitle[index], fontsize=8, horizontalalignment='center', \
verticalalignment='center', transform=figure.transFigure)
if cellnames and index==0:
bbox = subplot.get_position()
xloc = (bbox.xmax+bbox.xmin)/2.
yloc = (3*bbox.ymax+1)/4.
plot.text(xloc, yloc, cellnames[cellidx], fontsize=8, horizontalalignment='center', \
verticalalignment='bottom', transform=figure.transFigure)
if hist:
subplot = figure.add_subplot(numrows,numcols,(2*index+1)*numcols+cellidx+1)
subplot = remove_spines(subplot)
reads_unbound = np.power([read[0][idx].sum()+read[1][idx].sum() for idx in quantized[0] \
if read[0][idx].size==width and read[1][idx].size==width], 0.25)
reads_bound = np.power([read[0][idx].sum()+read[1][idx].sum() for idx in quantized[-1] \
if read[0][idx].size==width and read[1][idx].size==width], 0.25)
h0 = subplot.hist(reads_unbound, bins=200, color=colors[0], histtype='step', linewidth=0.2, normed=True)
h1 = subplot.hist(reads_bound, bins=200, color=colors[-1], histtype='step', linewidth=0.2, normed=True)
xmin = 0
xmax = max([reads_bound.max(), reads_unbound.max()])
ymin = 0
ymax = max([h0[0].max(), h1[0].max()])
subplot.axis([xmin, xmax, ymin, ymax])
for text in subplot.get_xticklabels():
text.set_fontsize(7)
text.set_verticalalignment('center')
ytick_locs = list(np.linspace(np.round(ymin,2),np.round(ymax,2),5))
ytick_labels = tuple(['%.2f'%s for s in ytick_locs])
subplot.set_yticks(ytick_locs)
subplot.set_yticklabels(ytick_labels, color='k', fontsize=6, horizontalalignment='right')
subplot.set_xlabel('Fourth root of total reads', fontsize=6, horizontalalignment='center')
legends = ['(%.2f,%.2f)'%(xt[quant].min(),xt[quant].max()) for quant in quantized]
leghandle = plot.figlegend(fwd, legends, loc='lower right', mode="expand", ncol=q)
for text in leghandle.texts:
text.set_fontsize(6)
leghandle.set_frame_on(False)
if title:
plot.suptitle(title, fontsize=10)
return figure
def plot_reads(Reads,
xts,
motiflen=None,
q=5,
quantile=True,
bounds=None,
subtitle=None,
cellnames=None,
title=None,
hist=False):
"""
reads is a list of list of binary tuples of lists of numpy array with read counts. The outermost
list is over different cell types. The next inner list
is over different TSS dist thresholds. The binary tuple is for same / opposite strands.
"""
width = Reads[0][0][0][0].size
figure = plot.figure()
xvals = np.arange(-width / 2, width / 2)
numcols = len(Reads)
numrows = len(Reads[0])
if hist:
numrows = 2 * numrows
colors = colorwheel(q)
for cellidx, reads in enumerate(Reads):
for index, (xt, read) in enumerate(zip(xts, reads)):
if quantile:
quantized = utils.quantiles(xt, q=q)
else:
quantized = utils.quantize(xt, q=q, bounds=bounds)
same = [
np.mean([
read[0][idx] for idx in quant if read[0][idx].size == width
], 0) for quant in quantized
]
opp = [
-1 * np.mean([
read[1][idx] for idx in quant if read[1][idx].size == width
], 0) for quant in quantized
]
if hist:
subplot = figure.add_subplot(numrows, numcols,
2 * index * numcols + cellidx + 1)
else:
subplot = figure.add_subplot(numrows, numcols,
index * numcols + cellidx + 1)
subplot = remove_spines(subplot)
fwd = [
subplot.plot(xvals, s, color=c, linestyle='-', linewidth=0.5)
for s, c in zip(same, colors)
]
rev = [
subplot.plot(xvals, o, color=c, linestyle='-', linewidth=0.5)
for o, c in zip(opp, colors)
]
subplot.axhline(0, linestyle='--', linewidth=0.2)
subplot.axvline(0, linestyle='--', linewidth=0.2)
if motiflen:
subplot.axvline(motiflen - 1,
linestyle='--',
c='g',
linewidth=0.2)
xmin = xvals[0]
xmax = xvals[-1]
ymax = max([s.max() for s in same])
ymin = min([o.min() for o in opp])
subplot.axis([xmin, xmax, ymin, ymax])
for text in subplot.get_xticklabels():
text.set_fontsize(7)
text.set_verticalalignment('center')
ytick_locs = list(
np.linspace(np.round(ymin, 2), np.round(ymax, 2), 5))
if 0 not in ytick_locs:
ytick_locs.append(0)
ytick_locs.sort()
ytick_labels = tuple(['%.2f' % s for s in ytick_locs])
subplot.set_yticks(ytick_locs)
subplot.set_yticklabels(ytick_labels,
color='k',
fontsize=6,
horizontalalignment='right')
if subtitle and cellidx == 0:
bbox = subplot.get_position()
xloc = bbox.xmin / 3.
yloc = (bbox.ymax + bbox.ymin) / 2.
plot.text(xloc, yloc, subtitle[index], fontsize=8, horizontalalignment='center', \
verticalalignment='center', transform=figure.transFigure)
if cellnames and index == 0:
bbox = subplot.get_position()
xloc = (bbox.xmax + bbox.xmin) / 2.
yloc = (3 * bbox.ymax + 1) / 4.
plot.text(xloc, yloc, cellnames[cellidx], fontsize=8, horizontalalignment='center', \
verticalalignment='bottom', transform=figure.transFigure)
if hist:
subplot = figure.add_subplot(
numrows, numcols, (2 * index + 1) * numcols + cellidx + 1)
subplot = remove_spines(subplot)
reads_unbound = np.power([read[0][idx].sum()+read[1][idx].sum() for idx in quantized[0] \
if read[0][idx].size==width and read[1][idx].size==width], 0.25)
reads_bound = np.power([read[0][idx].sum()+read[1][idx].sum() for idx in quantized[-1] \
if read[0][idx].size==width and read[1][idx].size==width], 0.25)
h0 = subplot.hist(reads_unbound,
bins=200,
color=colors[0],
histtype='step',
linewidth=0.2,
normed=True)
h1 = subplot.hist(reads_bound,
bins=200,
color=colors[-1],
histtype='step',
linewidth=0.2,
normed=True)
xmin = 0
xmax = max([reads_bound.max(), reads_unbound.max()])
ymin = 0
ymax = max([h0[0].max(), h1[0].max()])
subplot.axis([xmin, xmax, ymin, ymax])
for text in subplot.get_xticklabels():
text.set_fontsize(7)
text.set_verticalalignment('center')
ytick_locs = list(
np.linspace(np.round(ymin, 2), np.round(ymax, 2), 5))
ytick_labels = tuple(['%.2f' % s for s in ytick_locs])
subplot.set_yticks(ytick_locs)
subplot.set_yticklabels(ytick_labels,
color='k',
fontsize=6,
horizontalalignment='right')
subplot.set_xlabel('Fourth root of total reads',
fontsize=6,
horizontalalignment='center')
legends = [
'(%.2f,%.2f)' % (xt[quant].min(), xt[quant].max())
for quant in quantized
]
leghandle = plot.figlegend(fwd,
legends,
loc='lower right',
mode="expand",
ncol=q)
for text in leghandle.texts:
text.set_fontsize(6)
leghandle.set_frame_on(False)
if title:
plot.suptitle(title, fontsize=10)
return figure
