def pdf_for_model(sink, str_beta,mdot_low,mdot_hi,nsinks=64,corr=True): '''returns mdot pdf of 64*(tend-tbeg)/dt points following Peter's idea Returns: array of mdots for pdf, list of non nan indices of that arraydict of models, each model is a dict of the models it was ttest compared to ''' #time steady state begins from KS 2 samp stat tbeg= steady_state_begins(str_beta,units_tBH=True) tend= tbeg+3. if corr: sample= steady_rates(sink.mdot_vcorr,sink.tnorm(),tend-tbeg,tend,units_tBH=True,fill_nan=True) else: sample= steady_rates(sink.mdot,sink.tnorm(),tend-tbeg,tend,units_tBH=True,fill_nan=True) #check data is good assert( np.all(np.isfinite(sample)) ) #robust, either data exist in above range or not, will not have good data and nans in this sample #is desired, take subset of sinks sink_ids=np.arange(64) np.random.shuffle(sink_ids) sample= sample[ sink_ids[:nsinks] ] #can safely take log10 of rates now sample= np.log10(sample) #equal log spaced bins covering entire range of mdots bins= np.linspace(np.log10(mdot_low/2),np.log10(mdot_hi*2),num=41) # N bins = num-1 pdf,cdf= get_pdf_cdf(sample,bins) print 'sample= ',sample print 'bins= ',bins print 'pdf= ',pdf print 'cdf= ',cdf return sample,bins,pdf,cdf
def pdf_for_model_64_x_N(sink, str_beta,mdot_low,mdot_hi,nsinks=64,corr=True,last_2tBH=True,dt=2.,median=True): '''default: compute 64 point PDF taking median each sink over last 2 tBH can change dt = 1 to take 64x2 point PDF over last 2tBH or even dt=1 and last_2tBH=False to take 64xN point pdf between tsteady and tend of every model ''' if last_2tBH: tbeg= -2.+ min(19.8,sink.tnorm()[-1]) assert(tbeg >= steady_state_begins(str_beta,units_tBH=True)) else: tbeg= steady_state_begins(str_beta,units_tBH=True) #time steady state begins from KS 2 samp stat sample=[] for cnt in range(1,100): tend= tbeg+cnt*dt if tend > min(20.,sink.tnorm()[-1]): break if median: # Take median mdot over steady state time range for each sink if corr: sample+= list( steady_rates(sink.mdot_vcorr,sink.tnorm(),dt,tend,units_tBH=True,fill_nan=True,str_beta=str_beta,median=median) ) else: sample+= list( steady_rates(sink.mdot,sink.tnorm(),dt,tend,units_tBH=True,fill_nan=True,str_beta=str_beta,median=median) ) else: # Take mean if corr: sample+= list( steady_rates(sink.mdot_vcorr,sink.tnorm(),dt,tend,units_tBH=True,fill_nan=True,str_beta=str_beta,median=False) ) else: sample+= list( steady_rates(sink.mdot,sink.tnorm(),dt,tend,units_tBH=True,fill_nan=True,str_beta=str_beta,median=False) ) print 'inside loop pdf_for_model_64_x_N, dt= ',dt,'tend= ',tend sample= np.array(sample) #sanity check print 'finished pdf_for_model_64_x_N, model=%s, len(sample)=%d' % (str_beta,len(sample)) if last_2tBH and dt > 1.01: assert(len(sample) == 64) elif last_2tBH and dt < 1.01: assert(len(sample) == 128) else: assert(len(sample) >= 128) #check data is good assert( np.all(np.isfinite(sample)) ) #robust, either data exist in above range or not #can safely take log10 of rates now sample= np.log10(sample) #equal log spaced bins covering entire range of mdots bins= np.linspace(np.log10(mdot_low/2),np.log10(mdot_hi*2),num=41) # N bins = num-1 pdf,cdf= get_pdf_cdf(sample,bins) print 'bins[[0,-1]]= ',bins[[0,-1]] print 'pdf= ',pdf print 'cdf= ',cdf return sample,bins,pdf,cdf
def pdf_for_model_mean_of_many_pdfs(sink, str_beta,mdot_low,mdot_hi, tend=20.,dt=0.1): '''returns mdot pdf of 64*(tend-tbeg)/dt points following Peter's idea Returns: array of mdots for pdf, list of non nan indices of that arraydict of models, each model is a dict of the models it was ttest compared to ''' #corr mdots end_t,end_mach= t_rms_mach_near_end(str_beta) sink.rm_mdot_systematics(end_t,end_mach) #time steady state begins from KS 2 samp stat tbeg= steady_state_begins(str_beta,units_tBH=True) #median accretion rate in 0.1 bins between t=2tBH and t=end n_intervals= get_num_intervals(sink,tbeg,dt, tend=tend) pdfs= np.empty((n_intervals,64))*np.nan for ibin in range(n_intervals): #get 64 steady mdots for this time range pdfs[ibin,:]= steady_rates(sink.mdot_vcorr,sink.tnorm(),dt,tbeg+(ibin+1)*dt,units_tBH=True,fill_nan=True) #mask the array where have nans pdfs= ma.array(pdfs) pdfs.mask= np.isfinite(pdfs) == False #can safely take log10 of rates now pdfs= np.log10(pdfs) #equal log spaced bins covering entire range of mdots bins= np.linspace(np.log10(mdot_low/2),np.log10(mdot_hi*2),num=41) # N bins = num-1 #bins= bins[::-1] #start with smallest number first (most negative) print 'bins= ',bins #bin up data hist_vals= np.empty((n_intervals,len(bins)-1))*np.nan for i in range(n_intervals): if np.any(pdfs.mask[i,:]): continue hist_vals[i,:],b,c=plt.hist(pdfs[i,:],bins=bins) plt.close() hist_vals= ma.array(hist_vals) hist_vals.mask= np.isfinite(hist_vals) == False #compute averaged histogram final_hist= np.round(np.mean(hist_vals,axis=0)).astype(int) #average each bin count over n_intervals, round to nearest count, save as integer #convert histogram into array of values bin_c= (bins[1:]+bins[:-1])/2 final_arr=[] for cnt,c in enumerate(bin_c): final_arr+= [c]*final_hist[cnt] assert(len(final_arr) == final_hist.sum()) pdf,junk1,junk2= plt.hist(final_arr,bins=bins,normed=True) cdf,j1,j2=plt.hist(final_arr,bins=bins,cumulative=True,normed=True) plt.close() try: assert( ((bins[1:]-bins[:-1])*pdf).sum() == 1. ) except AssertionError: print '---- WARNING sum not 1, what is it? ------------ ' print '((bins[1:]-bins[:-1])*pdf).sum()= ',((bins[1:]-bins[:-1])*pdf).sum() #raise AssertionError #return array of values, bins, pdf and cdf return np.array(final_arr),bins,pdf,cdf
#raw_t= master_sink.old_t[-master_sink.tnorm().size:].copy()
print "%s" % args.str_beta
print "mdotTime, tnorm()"
for raw_t,norm in zip(master_sink.mdotTime,master_sink.tnorm()):
print raw_t,norm
print 'exiting early'
sys.exit()
#tend
print 'model: %s, last t/tBH= %.5f' % (args.str_beta,master_sink.tnorm()[-1])
if False: # print machrms at end
t_end= min(20.,master_sink.tnorm()[-1])
i_end= index_before_tBH(master_sink,t_end,units_tBH=True)
mach_end= master_sink.f_rms_mach(master_sink.mdotTime[i_end])
print 'model= %s, tsteady/tBH= %.2f, tend/tBH= %.2f, rms Mach end= %.2f' % \
(args.str_beta, steady_state_begins(args.str_beta,True),t_end,mach_end)
#corr mdots
end_t,end_mach= t_rms_mach_near_end(args.str_beta)
master_sink.rm_mdot_systematics(end_t,end_mach)
#final PDF
final_pdf= dict(arr={},bins={},pdf={},cdf={}) #PDF USING, 64 pdf, over last 2tBH
low,hi= low_hi_mdot_all_models()
final_pdf['arr'][args.str_beta],final_pdf['bins'][args.str_beta],final_pdf['pdf'][args.str_beta],final_pdf['cdf'][args.str_beta]= pdf_for_model_64_x_N(master_sink,args.str_beta, low,hi,nsinks=64,corr=True,last_2tBH=True,dt=2.)
#make plots
just_median(master_sink,args)
median_and_when_equil_begins(master_sink,args)
steady_pdf_cdf(master_sink,args)
all_rates(master_sink,args,corr=True)
bootstrap_sigma_vs_nsink(final_pdf,args.str_beta)
print 'finished'
"-str_beta",
nargs=7,
choices=["bInf", "bInfsd3", "b100", "b10", "b1", "b1e-1", "b1e-2"],
action="store",
help="directory containing parsed mdot files",
required=True,
)
parser.add_argument("-outdir", action="store", help="directory to save plots to", required=True)
parser.add_argument("-ylim", nargs=2, type=float, action="store", help="", required=False)
parser.add_argument("-xlim", nargs=2, type=float, action="store", help="", required=False)
args = parser.parse_args()
print "BEFORE CALC, fixed t_steady: "
for str_beta in args.str_beta:
tsteady = steady_state_begins(str_beta, True)
print "STEADY STATE: model=%s, t/tBH=%.2f, t/tABH=%.2f" % (str_beta, tsteady, tBH_to_tBHA(tsteady, str_beta))
sinks = {}
for fn, str_beta in zip(args.sinks, args.str_beta):
fin = open(fn, "r")
sinks[str_beta] = load(fin)
fin.close()
# corr mdots
end_t, end_mach = stats.t_rms_mach_near_end(str_beta)
sinks[str_beta].rm_mdot_systematics(end_t, end_mach)
# deterine min and max mdot of all models
low, hi = 100, 1.0e-10
for key in sinks.keys():
istart = stats.index_before_tBH(sinks[key], steady_state_begins(key, True))
iend = stats.index_before_tBH(sinks[key], 20.0)