def all_models_cdfs_pdfs2(cdf, pdf, bin_c, args, nsinks=64):
"""bins -- linear spaced 10^val not val"""
sns.set_style("whitegrid") # ,{"axes.facecolor": ".97"})
sns.set_palette("colorblind")
# kwargs
laba = dict(fontweight="bold", fontsize=dFS)
# str_beta map
addtext = {}
addtext["hydro_both"] = r"$\mathbf{ \beta_0 = \infty \, (1+2) }$"
addtext["b100"] = r"$\mathbf{ \beta_0 = 100 }$"
addtext["b10"] = r"$\mathbf{ \beta_0 = 10 }$"
addtext["b1"] = r"$\mathbf{ \beta_0 = 1 }$"
addtext["b1e-1"] = r"$\mathbf{ \beta_0 = 0.1 }$"
addtext["b1e-2"] = r"$\mathbf{ \beta_0 = 0.01 }$"
# plot
fig, axes = plt.subplots(2, 1, sharex=True)
ax = axes.flatten()
plt.subplots_adjust(hspace=0.1)
colors = sns.color_palette() # color blind colors b/c set_palette() called above
cnt = 0
for mod in cdf.keys():
if mod == "bInf" or mod == "bInfsd3":
continue
ax[1].plot(bin_c[mod] / mdotB(), cdf[mod], c=colors[cnt], label=addtext[mod])
ax[0].plot(bin_c[mod] / mdotB(), pdf[mod], c=colors[cnt], label=addtext[mod])
cnt += 1
# ax[1].legend(loc=4,fontsize=FS,frameon=True,fancybox=True)
# reorganize legend
handles, labels = ax[1].get_legend_handles_labels()
handles = handles[:2] + [handles[-1], handles[-2], handles[-3], handles[2]]
labels = labels[:2] + [labels[-1], labels[-2], labels[-3], labels[2]]
ax[1].legend(handles, labels, loc=(0.76, 0.01), ncol=1, fontsize=FS - 5, frameon=True, fancybox=True)
#
ylab = ax[1].set_ylabel("CDF", **laba)
ylab = ax[0].set_ylabel("PDF", **laba)
xlab = ax[1].set_xlabel(r"$\mathbf{\dot{M}/\dot{M}_{\rm{B}} }$", **laba)
# ax[1].set_xlim(-7,-2.5)
for i in range(2):
ax[i].set_xlim(1e-5, 1e-1)
ax[1].set_ylim(-0.1, 1.1)
ax[0].set_ylim([-0.1, ax[0].get_ylim()[-1]])
for i in range(2):
ax[i].set_xscale("log")
ax[i].tick_params(axis="both", labelsize=FS)
plt.savefig(
os.path.join(args.outdir, "all-models-cdfs-pdfs-nsinks-%d-2.png" % nsinks),
bbox_extra_artists=[xlab, ylab],
bbox_inches="tight",
dpi=150,
)
plt.close()
def mvp_for_peter(master_sink,args): #write MVP data to file for Peter rate=master_sink.mdot/mdotB() time=(master_sink.mdotTime-master_sink.mdotTime[0])/get_tBH() fout=open(os.path.join(os.path.dirname(args.master_sink),'%s.pickle' % args.str_beta),'w') master_sink=dump((rate,time),fout) fout.close()
def median_rate(ax,sink,args,raw=True,corr=True,clip=True,times=[],c_ls=[]): #tsteady):
laba=dict(fontweight='bold',fontsize='xx-large')
if raw: ax.plot(sink.tnorm(),np.median(sink.mdot,axis=0)/mdotB(),c='k',label='raw')
if corr: ax.plot(sink.tnorm(),np.median(sink.mdot_vcorr,axis=0)/mdotB(),c='b',label='corr')
if clip: ax.plot(sink.tnorm(),np.median(sink.mdot_vcorr_clip,axis=0)/mdotB(),c='r',label='corr, clip')
for time,cls in zip(times,c_ls):
ax.plot([time]*2,ax.get_ylim(),cls,lw=2.)
ax.set_xlabel(r'$\mathbf{ t/t_{\rm{BH}} }$',**laba)
#2nd axis
ax2 = ax.twiny()
ax2.set_xlabel(r'$\mathbf{ t/t_{\rm{ABH}} }$',fontweight='bold',fontsize='xx-large')
ax2.plot(tBH_to_tBHA(sink.tnorm(),args.str_beta),np.median(sink.mdot_vcorr,axis=0)/mdotB(),visible=False)
#ax.plot([t_steady]*2,ax[0].get_ylim(),'r--',label='Steady State')
ax.text(ax.get_xlim()[1]*0.01,ax.get_ylim()[1]*0.9,'Median',fontweight='bold',fontsize='large',va='center',ha='left')
ax.set_yscale('log')
if args.ylim: ax.set_ylim(args.ylim[0],args.ylim[1])
ax.legend(loc=4,fontsize='small')
ax.set_ylabel(r'Log $\mathbf{ \dot{M}/\dot{M}_{\rm{B}} }$',**laba)
def all_rates(sink, args, corr=True):
fig,ax=plt.subplots()
for i in range(sink.mdot.shape[0]):
if corr: ax.plot(sink.tnorm(),sink.mdot_vcorr[i,:]/mdotB()) #,label='ID: %d' % i)
else: ax.plot(sink.tnorm(),sink.mdot[i,:]/mdotB()) #,label='ID: %d' % i)
ax.set_xlabel(r'$\mathbf{ t/t_{\rm{BH}} }$',fontweight='bold',fontsize='xx-large')
#second x axis
ax2 = ax.twiny()
ax2.set_xlabel(r'$\mathbf{ t/t_{\rm{ABH}} }$',fontweight='bold',fontsize='xx-large')
ax2.plot(tBH_to_tBHA(sink.tnorm(),args.str_beta),sink.mdot_vcorr[i,:]/mdotB(),visible=False)
#y label, left vertical panels
ax.set_ylabel(r'Log $\mathbf{ \dot{M}/\dot{M}_B }$',fontweight='bold',fontsize='xx-large')
ax.set_yscale('log')
if corr: lab= 'Model: %s, corrected' % args.str_beta
else: lab= 'Model: %s' % args.str_beta
ax.text(ax.get_xlim()[1]*0.01,ax.get_ylim()[1]*0.9,lab,fontweight='bold',fontsize='large',ha='left',va='top')
#ax.legend(loc=(1.01,0),ncol=4) #,fontsize='small'
if corr: name='all-rates-corrected-%s.png' % args.str_beta
else: name='all-rates-%s.png' % args.str_beta
#ax.set_xlim(0,20)
plt.savefig(os.path.join(os.path.dirname(args.master_sink),name))
plt.close()
def plot_nmad(master,label):
kwargs= dict(labargs=dict(fontsize=16,fontweight='bold'),\
xtickargs=dict(fontsize=13),\
ytickargs=dict(fontsize=16),\
legargs=dict(fontsize=12))
matplotlib.rcParams['xtick.labelsize'] = kwargs['xtickargs']['fontsize']
matplotlib.rcParams['ytick.labelsize'] = kwargs['ytickargs']['fontsize']
#fig,axes=plt.subplots(1,3,sharey=True)
fig,ax= plt.subplots()
x,y= master.med/mdotB(), master.nmad/mdotB()
ax.scatter(x,y,s=50,c='b') #master.med/master.med.max())
ax.set_yscale('log')
ax.set_xscale('log')
ax.set_xlim(x.min()/2,x.max()*2)
ax.set_ylim(y.min()/2,y.max()*2)
if args.xlim and args.ylim:
ax.set_xlim(args.xlim[0],args.xlim[1])
ax.set_ylim(args.ylim[0],args.ylim[1])
ax.set_xlabel(r'Median($\dot{M}/\dot{M}_B$)',**kwargs['labargs'])
ax.set_ylabel(r'Med[|$\dot{M}$ - Med($\dot{M}$)|]/$\dot{M}_B$',**kwargs['labargs'])
ax.set_title(r'NMAD, $\beta_0$ = ' + label,**kwargs['labargs'])
plt.savefig('nmad_beta%s.png' % label)
plt.close()
def steady_rates(mdotData,timeData,steady_wid,steady_end,units_tBH=True,fill_nan=True,str_beta=None,median=True):
'''returns 64 length array of time averaged mdots for each sink
if data does not exist in steady_wid fills with np.nan'''
assert(units_tBH)
assert(mdotData.shape[0] == 64)
imin=np.where(timeData >= steady_end-steady_wid)[0].min()
imax=np.where(timeData <= steady_end)[0].max()
ind=range(imin,imax+1)
try:
if str_beta is not None:
# plot median and mean mdot on same panel
plt.plot(timeData,np.median(mdotData,axis=0)/mdotB(),'b-',label='median')
plt.plot(timeData,np.mean(mdotData,axis=0)/mdotB(),'g-',label='mean')
# Get bootstraps using mdot over sinks at each time pt
med= bootstrap2(mdotData[:,ind]/mdotB(),str_beta, on_median=True)
mean= bootstrap2(mdotData[:,ind]/mdotB(),str_beta, on_median=False)
xvals=[timeData[ind].min(),timeData[ind].max()]
plt.fill_between(xvals,med['low'],med['hi'],color='b',alpha=0.4)
plt.fill_between(xvals,mean['low'],mean['hi'],color='g',alpha=0.4)
print 'mdotData.shape=',mdotData.shape,'np.mean(mdotData,axis=0).shape=',np.mean(mdotData,axis=0).shape
plt.legend(loc=0)
plt.yscale('log')
plt.savefig('medmean_%s.png' % str_beta)
plt.close()
# Also store these to text file
fin=open('bootstrap_2.txt','a')
fin.write('#beta sample_med -booterr +booterr sample_mean -booterr +booterr\n')
fin.write('%s %.5g %.5g %.5g %.5g %.5g %.5g\n' % \
(str_beta,med['sample_med'],np.abs(med['med']-med['low']),np.abs(med['med']-med['hi']),\
mean['sample_mean'],np.abs(mean['med']-mean['low']),np.abs(mean['med']-mean['hi'])))
fin.close()
# indiv sink, median mean in steady state
#for sid in [0,10,20,30,55]:
# plt.plot(timeData[ind],mdotData[sid,ind]/mdotB(),'k-')
# xvals=[timeData[ind].min(),timeData[ind].max()]
# plt.plot(xvals,[np.median(mdotData[sid,ind])/mdotB()]*2,'b-',label='median')
# plt.plot(xvals,[np.mean(mdotData[sid,ind])/mdotB()]*2,'g-',label='mean')
# plt.legend(loc=0)
# plt.yscale('log')
# plt.savefig('medmean_sid%d_%s.png' % (sid,str_beta))
# plt.close()
# Make plots showing rms fluctuation for individual sink
#for sid in [0,10,20,30,55]:
# plt.plot(timeData[ind],mdotData[sid,ind]/mdotB(),'b-')
# xvals=[timeData[ind].min(),timeData[ind].max()]
# med= np.median(mdotData[sid,ind])/mdotB()
# std= np.std(mdotData[sid,ind]/mdotB() - med)
# plt.plot(xvals,[med]*2,'k--',label='median')
# plt.plot(xvals,[med+std]*2,'r--',label='+std')
# plt.plot(xvals,[med-std]*2,'r--',label='-std')
# plt.plot(xvals,[med+2*std]*2,'g--',label='+2std')
# plt.plot(xvals,[med-2*std]*2,'g--',label='-2std')
# plt.legend(loc=0)
# plt.yscale('log')
# plt.savefig('std_indivsink_sid%d_%s' % (sid,str_beta))
# plt.close()
# Compute median 1-sigma fluctuation for all sinks, bootstrap the errors
mdot= np.log10(mdotData[:,ind]/mdotB())
med= np.percentile(mdot,q=50.,axis=1)
sig_sink={}
for key,q in zip(['low','hi'],[15.8,84.2]):
sig_sink[key]= np.abs(med - np.percentile(mdot,q=q,axis=1))
assert(len(sig_sink[key]) == 64)
#sig_sink[key]= np.median(sig_sink[key])
#one_sigma= 1.*np.std(mdot.T - med,axis=0)
# take median of one_sigmas and bootstrap it for error
#med= np.median(one_sigma)
#d_boots,final,std_pdf= bootstrap(one_sigma,str_beta,print_results=False)
print "INDIV SINK MEDIAN 1-SIGMA = percentile(log10(mdot/mdotB)) then median over sinks: $median^{+q75}_{-q25}$"
print "%s: $^{+%.8f}_{-%.8f}$" % (str_beta,np.median(sig_sink['hi']),np.median(sig_sink['low']))
for sid in [0,4,20,42,55,63]:
plt.plot(timeData[ind],np.log10(mdotData[sid,ind]/mdotB()),'b-')
xvals=[timeData[ind].min(),timeData[ind].max()]
plt.fill_between(xvals,med[sid]-sig_sink['low'][sid],med[sid]+sig_sink['hi'][sid],color='b',alpha=0.4)
plt.savefig('onesig_sid-%d_%s.png' % (sid,str_beta))
plt.close()
#a,b,c=plt.hist(two_sigmas)
#plt.plot([med]*2,[0,np.max(a)],'b--',label='median')
#plt.plot([final['med']-final['med_low']]*2,[0,np.max(a)],'r--',label='- err median')
#plt.plot([final['med']+final['med_hi']]*2,[0,np.max(a)],'r--',label='+ err median')
#plt.legend(loc=0)
#plt.savefig('twosigma_indivsink_%s' % str_beta)
#plt.close()
if median:
return np.median(mdotData[:,ind],axis=1)
else:
print '----WARNING----: MEAN over steady state, bool_median=',median
return np.mean(mdotData[:,ind],axis=1)
except IndexError:
print 'WARNING: no data between %.1f < t/tBH < %.1f' % (steady_end-steady_wid,steady_end)
return np.empty(mdotData.shape[0])*np.nan
def errorbars_vs_beta2(final64,use_rms_beta=True,name='test.png'):
'''all mdot quantities are in units of mdot/mdotB'''
sns.set_style('ticks') #,{"axes.facecolor": ".97"})
sns.set_palette('colorblind')
c_med='k' #median
c_mean= sns.color_palette()[0] #'b'
c_bhline= sns.color_palette()[1] #'g'
c_aaron=sns.color_palette()[2] #error on median box 'r'
c_mark= sns.color_palette()[3] #'pink'
#kwargs
laba=dict(fontweight='bold',fontsize=dFS)
kwargs_axtext=dict(fontweight='bold',fontsize=FS,va='top')
# orig final measure, REPLACED BY bootstrap_PDF.txt
#med,med_low,med_hi,std_low,std_hi,yax_val,rms_beta,med_beta,name= put_errbars_in_arrays_linear(final64)
# MEDIAN, bootstrap PDF
bPDF= get_measurment('boot_PDF')
bPDF= add_to_measurement(bPDF)
# MEAN, bootstrap 2
b2= get_measurment('boot_2')
b2= add_to_measurement(b2)
# get rms_beta
rms_beta= bPDF['rms_beta']
#marks results
rmdot0= 0.0166
# Values when submitted to MNRAS
#mark_norm_mdot0=dict(median_sim=0.35,median_pred_256=0.36,err_median_pred_256=0.1,\
# mean_pred_256=1.26,err_mean_pred_256=0.1,\
# mean_sim=0.9,err_mean_sim=0.1) # 0.1 is guess
# Values for revision, after fitting marks table 1
mark_norm_mdot0=dict(median_sim=0.35,median_pred_256=0.36,err_median_pred_256=0.106,\
mean_pred_256=1.26,err_mean_pred_256=0.100,\
mean_sim=0.9,err_mean_sim=0.1) # 0.1 is guess
mark_norm_mdotB={}
for key in mark_norm_mdot0.keys():
mark_norm_mdotB[key]= mark_norm_mdot0[key]* 4/np.exp(1.5)/5.**3
#special figure for different sized subplots
fig = plt.figure(figsize=(10,6))
gs = gridspec.GridSpec(1, 2, width_ratios=[3, 1])
fig.subplots_adjust(wspace=0.05)
ax=[0,0]
ax[0] = plt.subplot(gs[0])
ax[1] = plt.subplot(gs[1])
#B != 0 points go on left side
#ax[0].errorbar(rms_beta[:-2],med[:-2]/mdotB(),yerr=[med_low[:-2]/mdotB(),med_hi[:-2]/mdotB()], fmt='o',ms=6,mew=2.,mfc='none',mec=c_med,c=c_med,label=r'This Study (Median)')
# My Magnetic Points
# median
#ax[0].errorbar(rms_beta[:-2],med[:-2]/mdotB(),yerr=[med_low[:-2]/mdotB(),med_hi[:-2]/mdotB()], fmt='o',ms=6,mew=2.,mfc='none',mec=c_mark,c=c_mark,label='Krumholz et al. (2006)')
ax[0].errorbar(bPDF['rms_beta'][:-2],bPDF['med'][:-2],yerr=[bPDF['med_low'][:-2],bPDF['med_hi'][:-2]], fmt='o',ms=6,mew=2.,mfc='none',mec=c_mark,c=c_mark,label='Krumholz et al. (2006)')
ax[0].errorbar(bPDF['rms_beta'][:-2],bPDF['med'][:-2],yerr=[bPDF['med_low'][:-2],bPDF['med_hi'][:-2]], fmt='o',ms=6,mew=2.,mfc='none',mec=c_med,c=c_med,label='Median')
#ax[0].errorbar(b2['rms_beta'][:-2],b2['med'][:-2],yerr=[b2['med_low'][:-2],b2['med_hi'][:-2]], fmt='o',ms=6,mew=2.,mfc='none',mec='b',c='b',label='boot 2 median')
# mean
ax[0].errorbar(b2['rms_beta'][:-2],b2['mean'][:-2],yerr=[b2['mean_low'][:-2],b2['mean_hi'][:-2]], fmt='o',ms=6,mew=2.,mfc='none',mec=c_mean,c=c_mean,label='Mean')
#just for legend, plot some junk with right colors for mine and marks
#ax[0].plot([5,6],med[-2:],c=c_med,visible=False,label=r'This Study (Median $\mathbf{Log_{\rm{10}} \,\, \dot{M} }$)')
#ax[0].plot([5,6],med[-2:],c=c_mark,visible=False,label='Krumholz et al. (2006)')
#aarons
cont_beta=np.logspace(-3,1,num=50)
ax[0].plot(cont_beta,lee_parallel(cont_beta,b_norm=True),c=c_aaron,ls='--',lw=2,label='Lee et al. (2014)')
#print useful number for paper
#print 'ratio of mdot to (mdot_perp+mdot_par)/2 is: betarms,mdot,mdotpar,mdot/mdotpar'
#for myb,mymdot,mdotpar in zip(rms_beta[:-2],med[:-2],np.log10(lee_parallel(rms_beta[:-2]))): print myb,mymdot,mdotpar,10**(mdotpar)/10**(mymdot)
#mdot BH
ax[0].plot(ax[0].get_xlim(),[mdotBH()/mdotB()]*2,c=c_bhline,ls='--',lw=2)
ax[0].text(1e-2,mdotBH()/mdotB(),r'$\mathbf{ \dot{M}_{BH} }$',verticalalignment='bottom',**text) #,transform=ax[0].transAxes)
#ax[0].text(1e-2,np.log10(mdotBH()),r'$\mathbf{ \dot{M}_{BH} }$',color=c_bhline,**kwargs_ax[0].ext)
#B = 0 points go on right side
ax[1].plot(range(10),visible=False)
#Marks prediction
ax[1].errorbar(4,mark_norm_mdotB['median_pred_256'],yerr=mark_norm_mdotB['err_median_pred_256'],fmt='o',ms=6,mew=2.,mfc='none',mec=c_mark,c=c_mark,label='Krumholz et al. (2006)')
#ax[1].errorbar(4,mark_norm_mdotB['mean_pred_256'],yerr=mark_norm_mdotB['err_mean_pred_256'],fmt='o',ms=6,mew=2.,mfc='none',mec=c_mark,c=c_mark,label='Krumholz et al. (2006)')
ax[1].errorbar(4,mark_norm_mdotB['mean_sim'],yerr=mark_norm_mdotB['err_mean_sim'],fmt='o',ms=6,mew=2.,mfc='none',mec=c_mark,c=c_mark,label='Krumholz et al. (2006)')
#my hydro pts
#ax[1].errorbar([5,6],med[-2:]/mdotB(),yerr=[med_low[-2:]/mdotB(),med_hi[-2:]/mdotB()], fmt='o',ms=6,mew=2.,mfc='none',mec=c_med,c=c_med)
ax[1].errorbar([5,6],bPDF['med'][-2:],yerr=[bPDF['med_low'][-2:],bPDF['med_hi'][-2:]], fmt='o',ms=6,mew=2.,mfc='none',mec=c_med,c=c_med)
#ax[1].errorbar([9,10],b2['med'][-2:],yerr=[b2['med_low'][-2:],b2['med_hi'][-2:]], fmt='o',ms=6,mew=2.,mfc='none',mec='r',c='r')
ax[1].errorbar([5,6],b2['mean'][-2:],yerr=[b2['mean_low'][-2:],b2['mean_hi'][-2:]], fmt='o',ms=6,mew=2.,mfc='none',mec=c_mean,c=c_mean)
#mdotBH and same for hydro limit of lee14
ax[1].plot(ax[1].get_xlim(),[mdotBH()/mdotB()]*2,c=c_bhline,ls='--',lw=2)
ax[1].plot([0.5,8.5],[mdotBH()/mdotB()]*2,c=c_aaron,ls='--',lw=2)
#finish labeling
for i in range(2):
ax[i].set_ylim(6e-4,1e-2)
ax[0].set_xscale('log')
ax[0].set_xlim(8e-3,1e1)
#ax[1].set_xlim(2,1e1)
ylab= ax[0].set_ylabel(r'$\mathbf{\dot{M}/\dot{M}_{\rm{B}} }$',**laba)
xlab= fig.text(0.5, 0., r'$\mathbf{ \beta_{turb} }$', ha='center', **laba) #ax[0].set_xlabel(r'$\mathbf{ \beta_{rms} }$',**laba)
ax[0].legend(loc=4,fontsize=FS,frameon=True,fancybox=True)
#reorganize legend
handles, labels = ax[0].get_legend_handles_labels()
handles= [handles[-1],handles[-2],handles[1],handles[0]]
labels= [labels[-1],labels[-2],labels[1],labels[0]]
ax[0].legend(handles, labels,loc=4,frameon=True,fancybox=True,fontsize=FS)
#remove top and right axis lines
#sns.despine()
#remove spines and ticks between ax and ax2
#ax[0].spines['right'].set_visible(False)
#ax[1].spines['left'].set_visible(False)
for i in range(2): ax[i].tick_params(axis='both', which='major', labelsize=FS)
ax[1].set_yticks([])
#add slashes indicating break in axis
d = .015 # how big to make the diagonal lines in axes coordinates
# arguments to pass plot, just so we don't keep repeating them
kwargs = dict(transform=ax[0].transAxes, color='k', clip_on=False)
ax[0].plot((1-d,1+d), (-d,+d), **kwargs)
ax[0].plot((1-d,1+d),(1-d,1+d), **kwargs)
kwargs.update(transform=ax[1].transAxes) # switch to the bottom axes
ax[1].plot((-d*3,+d*3), (1-d,1+d), **kwargs)
ax[1].plot((-d*3,+d*3), (-d,+d), **kwargs)
#put infinity on 2nd axis
ax[1].set_xticks([])
ax[1].set_xlabel(r'$\mathbf{ \infty }$',**laba)
#plt.tick_params(axis='both', which='major', labelsize='large')
# show only some of yaxis ticks
#ax[0].yaxis.tick_left()
ax[1].yaxis.tick_right()
# y log scale
for i in range(2): ax[i].set_yscale('log')
#SAVE IS BROKEN, save manually
plt.tight_layout()
#plt.show()
#save
name='testing.png'
plt.savefig(name, bbox_extra_artists=[xlab,ylab], bbox_inches='tight',dpi=150)
print 'wrote %s' % name
def errorbars_vs_beta2(final64, use_rms_beta=True):
"""all mdot quantities are in units of mdot/mdotB"""
sns.set_style("ticks") # ,{"axes.facecolor": ".97"})
sns.set_palette("colorblind")
c_med = "k" # median
c_bhline = sns.color_palette()[0] #'b'
c_aaron = sns.color_palette()[2] # error on median box 'r'
c_mark = sns.color_palette()[1] #'g'
# kwargs
laba = dict(fontweight="bold", fontsize="xx-large")
kwargs_axtext = dict(fontweight="bold", fontsize="x-large", va="top")
# final measure
med, med_low, med_hi, std_low, std_hi, yax_val, rms_beta, med_beta, name = put_errbars_in_arrays_linear(final64)
# marks results
rmdot0 = 0.0166
mark_norm_mdot0 = dict(median_sim=0.35, median_pred_256=0.36, err_median_pred_256=0.1)
mark_norm_mdotB = {}
for key in mark_norm_mdot0.keys():
mark_norm_mdotB[key] = mark_norm_mdot0[key] * 4 / np.exp(1.5) / 5.0 ** 3
# special figure for different sized subplots
fig = plt.figure()
gs = gridspec.GridSpec(1, 2, width_ratios=[3, 1])
fig.subplots_adjust(wspace=0.05)
ax = [0, 0]
ax[0] = plt.subplot(gs[0])
ax[1] = plt.subplot(gs[1])
# B != 0 points go on left side
ax[0].errorbar(
rms_beta[:-2],
med[:-2] / mdotB(),
yerr=[med_low[:-2] / mdotB(), med_hi[:-2] / mdotB()],
fmt="o",
ms=6,
mew=2.0,
mfc="none",
mec=c_mark,
c=c_mark,
label="Krumholz et al. (2006)",
)
ax[0].errorbar(
rms_beta[:-2],
med[:-2] / mdotB(),
yerr=[med_low[:-2] / mdotB(), med_hi[:-2] / mdotB()],
fmt="o",
ms=6,
mew=2.0,
mfc="none",
mec=c_med,
c=c_med,
label=r"This Study (Median)",
)
# just for legend, plot some junk with right colors for mine and marks
# ax[0].plot([5,6],med[-2:],c=c_med,visible=False,label=r'This Study (Median $\mathbf{Log_{\rm{10}} \,\, \dot{M} }$)')
# ax[0].plot([5,6],med[-2:],c=c_mark,visible=False,label='Krumholz et al. (2006)')
# aarons
cont_beta = np.logspace(-3, 1, num=50)
ax[0].plot(cont_beta, lee_parallel(cont_beta, b_norm=True), c=c_aaron, ls="--", lw=2, label="Lee et al. (2014)")
# print useful number for paper
print "ratio of mdot to (mdot_perp+mdot_par)/2 is: betarms,mdot,mdotpar,mdot/mdotpar"
for myb, mymdot, mdotpar in zip(rms_beta[:-2], med[:-2], np.log10(lee_parallel(rms_beta[:-2]))):
print myb, mymdot, mdotpar, 10 ** (mdotpar) / 10 ** (mymdot)
# mdot BH
ax[0].plot(
ax[0].get_xlim(), [mdotBH() / mdotB()] * 2, c=c_bhline, ls="--", lw=2, label=r"$\mathbf{ \dot{M}_{BH} }$"
)
# ax[0].text(1e-2,np.log10(mdotBH()),r'$\mathbf{ \dot{M}_{BH} }$',color=c_bhline,**kwargs_ax[0].ext)
# B = 0 points go on right side
ax[1].plot(range(10), visible=False)
# Marks prediction
ax[1].errorbar(
4,
mark_norm_mdotB["median_pred_256"],
yerr=mark_norm_mdotB["err_median_pred_256"],
fmt="o",
ms=6,
mew=2.0,
mfc="none",
mec=c_mark,
c=c_mark,
label="Krumholz et al. (2006)",
)
# my hydro pts
ax[1].errorbar(
[5, 6],
med[-2:] / mdotB(),
yerr=[med_low[-2:] / mdotB(), med_hi[-2:] / mdotB()],
fmt="o",
ms=6,
mew=2.0,
mfc="none",
mec=c_med,
c=c_med,
)
# mdotBH and same for hydro limit of lee14
ax[1].plot(ax[1].get_xlim(), [mdotBH() / mdotB()] * 2, c=c_bhline, ls="--", lw=2)
ax[1].plot([0.5, 8.5], [mdotBH() / mdotB()] * 2, c=c_aaron, ls="--", lw=2)
# finish labeling
for i in range(2):
ax[i].set_ylim(6e-4, 1e-2)
ax[0].set_xscale("log")
ax[0].set_xlim(8e-3, 1e1)
# ax[1].set_xlim(2,1e1)
ylab = ax[0].set_ylabel(r"$\mathbf{\dot{M}/\dot{M}_{\rm{B}} }$", **laba)
xlab = fig.text(
0.5, 0.0, r"$\mathbf{ \beta_{rms} }$", ha="center", **laba
) # ax[0].set_xlabel(r'$\mathbf{ \beta_{rms} }$',**laba)
ax[0].legend(loc=4, fontsize="medium", frameon=True, fancybox=True)
# reorganize legend
handles, labels = ax[0].get_legend_handles_labels()
handles = [handles[-1], handles[-2], handles[0], handles[1]]
labels = [labels[-1], labels[-2], labels[0], labels[1]]
ax[0].legend(handles, labels, loc=4, fontsize="medium", frameon=True, fancybox=True)
# remove top and right axis lines
# sns.despine()
# remove spines and ticks between ax and ax2
# ax[0].spines['right'].set_visible(False)
# ax[1].spines['left'].set_visible(False)
for i in range(2):
ax[i].tick_params(axis="both", which="major", labelsize="large")
ax[1].set_yticks([])
# add slashes indicating break in axis
d = 0.015 # how big to make the diagonal lines in axes coordinates
# arguments to pass plot, just so we don't keep repeating them
kwargs = dict(transform=ax[0].transAxes, color="k", clip_on=False)
ax[0].plot((1 - d, 1 + d), (-d, +d), **kwargs)
ax[0].plot((1 - d, 1 + d), (1 - d, 1 + d), **kwargs)
kwargs.update(transform=ax[1].transAxes) # switch to the bottom axes
ax[1].plot((-d * 3, +d * 3), (1 - d, 1 + d), **kwargs)
ax[1].plot((-d * 3, +d * 3), (-d, +d), **kwargs)
# put infinity on 2nd axis
ax[1].set_xticks([])
ax[1].set_xlabel(r"$\mathbf{ \infty }$", fontsize="large")
# plt.tick_params(axis='both', which='major', labelsize='large')
# show only some of yaxis ticks
# ax[0].yaxis.tick_left()
ax[1].yaxis.tick_right()
# y log scale
for i in range(2):
ax[i].set_yscale("log")
# save
plt.savefig(
os.path.join(args.outdir, "errorbars_v_brms2.png"), bbox_extra_artists=[xlab, ylab], bbox_inches="tight"
)
def model_vs_errorbars2(final64):
"""final64 is linear mdots now
put model names on y axis, error bars on x axis"""
sns.set_style("ticks") # ,{"axes.facecolor": ".97"})
sns.set_palette("colorblind")
c_med = "k" # median
c_sem = sns.color_palette()[2] # error on median box 'r', bootstraped
c_std = sns.color_palette()[0] # std dev of pdf 'b', percentiled
# kwargs
laba = dict(fontweight="bold", fontsize="xx-large")
kwargs_axtext = dict(fontweight="bold", fontsize="x-large", va="top", ha="left")
# mapping
addtext = {}
addtext["bInf"] = r"$\mathbf{ \beta_0 = \infty }$ (1)"
addtext["bInfsd3"] = r"$\mathbf{ \beta_0 = \infty }$ (2)"
addtext["b100"] = r"$\mathbf{ \beta_0 = 100 }$"
addtext["b10"] = r"$\mathbf{ \beta_0 = 10 }$"
addtext["b1"] = r"$\mathbf{ \beta_0 = 1 }$"
addtext["b1e-1"] = r"$\mathbf{ \beta_0 = 0.1 }$"
addtext["b1e-2"] = r"$\mathbf{ \beta_0 = 0.01 }$"
# plot
fig, ax = plt.subplots()
ax.set_ylim(0, 8)
ax.set_xlim(1e-4, 2e-2)
ylabels = [item.get_text() for item in ax.get_yticklabels()] # y tick labels
meds, med_lows, med_his, std_lows, std_his, yax_vals, rms_betas, med_betas, names = put_errbars_in_arrays_linear(
final64
)
for med, med_low, med_hi, std_low, std_hi, yax_val, rms_beta, med_beta, name in zip(
meds, med_lows, med_his, std_lows, std_his, yax_vals, rms_betas, med_betas, names
):
if name == names[-1]:
box_and_whisker(
ax,
med / mdotB(),
(med_low / mdotB(), med_hi / mdotB()),
(std_low / mdotB(), std_hi / mdotB()),
yax_val,
legend_lab_ea_color=True,
)
else:
box_and_whisker(
ax, med / mdotB(), (med_low / mdotB(), med_hi / mdotB()), (std_low / mdotB(), std_hi / mdotB()), yax_val
)
ylabels[yax_val] = addtext[name]
ax.set_yticklabels(ylabels, fontsize="large")
xlab = ax.set_xlabel(r"$\mathbf{\dot{M}/\dot{M}_{\rm{B}} }$", **laba)
ylab = ax.set_ylabel("Simulation", **laba)
ax.legend(loc=4, fontsize=10, frameon=True, fancybox=True)
plt.tick_params(axis="x", which="major", labelsize="large")
# turn off horizontal grid lines
ax.yaxis.grid(False)
ax.xaxis.grid(True)
ax.set_xscale("log")
ax.yaxis.tick_left()
# ax.set_yticks([])
# remove top and right axis lines
# sns.despine()
# save
plt.savefig(
os.path.join(args.outdir, "models-vs-errorbars2.png"), bbox_extra_artists=[xlab, ylab], bbox_inches="tight"
)
raise ValueError
model_v_model_cdfs_pdfs(final_pdf["arr"], final_pdf["bins"], final_pdf["cdf"], final_pdf["pdf"], args, nsinks=nsinks)
# final measurement
# print 'Quantities using mdot= Log10(Mdot)'
# final_meas= stats.final_measurement(final_pdf['arr'])
# This gives correct Mdot/MdotB final measurements BUT the values CAN be computed from above log10(Mdot) like this
# q25,q50,q75 = above q25,q50,q75 / mdotB
# define: y= log10(mdot), x= mdot/mdotB
# then x = exp(y*ln10)/mdotB
# so sigma_x = sigma_y* exp(y*ln10)*ln10/mdotB)
# doing that agrees with numbers output below
final_pdf_linear, final_pdf_log = {}, {}
for key in final_pdf["arr"].keys():
# final_pdf_linear[key]= np.log10( np.power(10,final_pdf['arr'][key])/ mdotB() )
final_pdf_linear[key] = np.power(10, final_pdf["arr"][key]) / mdotB()
final_pdf_log[key] = np.log10(np.power(10, final_pdf["arr"][key]) / mdotB())
print "Quantities using mdot/mdotB, where mdotB=", mdotB()
final_meas_linear = stats.final_measurement(final_pdf_linear)
print "Quantities using log10(mdot/mdotB), where mdotB=", mdotB()
final_meas_log = stats.final_measurement(final_pdf_log)
exit("exiting after print final mean measurements")
# The plots below used to work when mdot = mdot without mdotB normalization
# They don't know but can recreate by going to appropriate commmit #
for key in final_pdf_linear.keys():
final_pdf_linear[key] = final_pdf_linear[key] * mdotB()
# plot model name vs. errobars, do for both final64 and final128
model_vs_errorbars(final_meas)
master={}
for sink_dir,str_beta in zip(args.dirs,args.labels):
fin=open(os.path.join(sink_dir,'master_sink.pickle'),'r')
master[str_beta]=pickle.load(fin)
master[str_beta].sink_xyz(args.init_sink)
fin.close()
sizes=range(len(args.labels))
for i,str_beta in enumerate(args.labels):
sizes[i]= master[str_beta].mdotTime.shape[0]
junk=args.labels[0]
x= np.empty( (master[junk].mdot.shape[0],)+(max(sizes),)+(len(args.labels),))*np.nan
y=x.copy()
for i,str_beta in enumerate(args.labels):
time=(master[str_beta].mdotTime-master[str_beta].mdotTime[0])/get_tBH()
x[:,:sizes[i],i]= np.transpose(np.repeat(time[:,np.newaxis], x.shape[0], axis=1)) #(npts,) array to (64,npts,)
y[:,:sizes[i],i]= master[str_beta].mdot/mdotB()
titles= np.core.defchararray.add(np.array(['sid ']*len(master[str_beta].init.sid)),master[str_beta].init.sid.astype(str))
nrow,ncol=8,8
multi_plot(nrow,ncol,x,y,titles=titles,logy=True,ylim=args.ylim)
print 'exiting after correlated sinks'
sys.exit(0)
#nmad plot
print 'plotting nmad'
for sink_dir,label in zip(args.dirs,args.labels):
fin=open(os.path.join(sink_dir,'master_sink.pickle'),'r')
master=pickle.load(fin)
fin.close()
plot_nmad(master,label)
