def fithist(hist,central,eff,plotDir):
data = []
weightdata = []
for i in range(hist.GetXaxis().GetNbins()):
data.append(hist.GetBinCenter(i))
weightdata.append(hist.GetBinContent(i)) #weight according to height
data = array(data)
weightdata = array(weightdata)
n,bins,patches = plt.hist(data,normed=True,bins=100,weights=weightdata,facecolor='b',histtype='stepfilled')
weightdata = weightdata/sum(weightdata)
if central == 'absolute_median' or central == 'mode':
if eff<0 or eff>1: raise RuntimeError('In order to use absolute IQR, you have to provide the reconstruction efficiency. Use --eff option.')
(mu,mu_err,sigma,sigma_err,upper_quantile,lower_quantile,err) = distribution_values(data,weightdata,central,eff=eff)
plt.plot((mu,mu),(0,plt.ylim()[1]),'r--',linewidth=2)
height = 0.607*max(n) #height at x=1*sigma in normal distribution
if lower_quantile>float('-inf'):
plt.plot((lower_quantile,upper_quantile),(height,height),'r--',linewidth=2)
plt.plot((lower_quantile,lower_quantile),(height-0.02,height+0.02),'r-',linewidth=2)
else:
plt.plot((mu,upper_quantile),(height,height),'r--',linewidth=2)
plt.plot((upper_quantile,upper_quantile),(height-0.02,height+0.02),'r-',linewidth=2)
if central == 'median':
(mu,mu_err,sigma,sigma_err,upper_quantile,lower_quantile) = distribution_values(data,weightdata,central)
plt.plot((mu,mu),(0,plt.ylim()[1]),'r--',linewidth=2)
height = 0.607*max(n) #height at x=1*sigma in normal distribution
plt.plot((lower_quantile,upper_quantile),(height,height),'r--',linewidth=2)
plt.plot((lower_quantile,lower_quantile),(height-0.02,height+0.02),'r-',linewidth=2)
plt.plot((upper_quantile,upper_quantile),(height-0.02,height+0.02),'r-',linewidth=2)
if central == 'mean':
(mu,mu_err,sigma,sigma_err) = distribution_values(data,weightdata,central)
gfunc = norm
y = gfunc.pdf( bins, mu, sigma)
plt.plot((mu,mu),(0,gfunc.pdf(mu,mu,sigma)),'r--',linewidth=2)
l = plt.plot(bins, y, 'r--', linewidth=2)
if central == 'trimmed':
(mu,mu_err,sigma,sigma_err,lower,upper) = distribution_values(data,weightdata,central)
#print mu,sigma,ptbin
gfunc = norm
y = gfunc.pdf(bins, mu, sigma)
plt.plot((mu,mu),(0,gfunc.pdf(mu,mu,sigma)),'r--',linewidth=2)
newbins = bins[all([bins>lower,bins<upper],axis=0)]
newy = y[all([bins>lower,bins<upper],axis=0)]
l = plt.plot(newbins, newy, 'r--', linewidth=2)
plt.xlabel('$p_T^{reco}/p_T^{true}$')
plt.ylabel('a.u.')
plt.savefig(plotDir+'/'+'histfit_'+central+'.png')
plt.close()
#avgres.append(mu)
#avgres_errs.append(mu_err)
#sigmaRs.append(sigma)
#sigmaR_errs.append(sigma_err)
return mu,mu_err,sigma,sigma_err
def fithist(data,weightdata,central,eff=eff,plotDir=None):
plotting = not (plotDir==None)
if plotting and not os.path.exists(plotDir):
print '== Making folder '+options.plotDir+' =='
os.makedirs(options.plotDir)
if plotting:
try:
from rootpy.plotting.style import set_style, get_style
print '== Using ATLAS style =='
atlas = get_style('ATLAS')
atlas.SetPalette(51)
set_style(atlas)
set_style('ATLAS',mpl=True)
except ImportError: print '== Not using ATLAS style (Can\'t import rootpy.) =='
if plotting: n,bins,patches = plt.hist(data,normed=True,bins=100,weights=weightdata,facecolor='b',histtype='stepfilled')
weightdata = weightdata/sum(weightdata)
if central == 'absolute_median' or central == 'mode':
if eff<0 or eff>1: raise RuntimeError('In order to use absolute IQR, you have to provide the reconstruction efficiency. Use --eff option.')
(mu,mu_err,sigma,sigma_err,upper_quantile,lower_quantile,err) = distribution_values(data,weightdata,central,eff=eff)
if plotting:
plt.plot((mu,mu),(0,plt.ylim()[1]),'r--',linewidth=2)
height = 0.607*max(n) #height at x=1*sigma in normal distribution
if lower_quantile>float('-inf'):
plt.plot((lower_quantile,upper_quantile),(height,height),'r--',linewidth=2)
plt.plot((lower_quantile,lower_quantile),(height-0.02,height+0.02),'r-',linewidth=2)
else:
plt.plot((mu,upper_quantile),(height,height),'r--',linewidth=2)
plt.plot((upper_quantile,upper_quantile),(height-0.02,height+0.02),'r-',linewidth=2)
if central == 'median':
(mu,mu_err,sigma,sigma_err,upper_quantile,lower_quantile) = distribution_values(data,weightdata,central)
if plotting:
plt.plot((mu,mu),(0,plt.ylim()[1]),'r--',linewidth=2)
height = 0.607*max(n) #height at x=1*sigma in normal distribution
plt.plot((lower_quantile,upper_quantile),(height,height),'r--',linewidth=2)
plt.plot((lower_quantile,lower_quantile),(height-0.02,height+0.02),'r-',linewidth=2)
plt.plot((upper_quantile,upper_quantile),(height-0.02,height+0.02),'r-',linewidth=2)
if central == 'mean':
(mu,mu_err,sigma,sigma_err) = distribution_values(data,weightdata,central)
if plotting:
gfunc = norm
y = gfunc.pdf( bins, mu, sigma)
plt.plot((mu,mu),(0,gfunc.pdf(mu,mu,sigma)),'r--',linewidth=2)
l = plt.plot(bins, y, 'r--', linewidth=2)
if central == 'trimmed':
(mu,mu_err,sigma,sigma_err,lower,upper) = distribution_values(data,weightdata,central)
if plotting:
#print mu,sigma,ptbin
gfunc = norm
y = gfunc.pdf(bins, mu, sigma)
plt.plot((mu,mu),(0,gfunc.pdf(mu,mu,sigma)),'r--',linewidth=2)
newbins = bins[all([bins>lower,bins<upper],axis=0)]
newy = y[all([bins>lower,bins<upper],axis=0)]
l = plt.plot(newbins, newy, 'r--', linewidth=2)
if plotting:
plt.xlabel('$p_T^{reco}/p_T^{true}$')
plt.ylabel('a.u.')
plt.savefig(plotDir+'/'+'histfit_'+central+'.png')
plt.close()
#avgres.append(mu)
#avgres_errs.append(mu_err)
#sigmaRs.append(sigma)
#sigmaR_errs.append(sigma_err)
return mu,mu_err,sigma,sigma_err