def run_EOS(EOS, m1, m2, thetas, type_set = 'None', N_EOS = 100, model_set = 'Bu2019inc', chirp_q = False):
chi = 1
if type_set == 'None':
sys.exit('Type is not defined')
num_samples = N_EOS
if type_set == 'BNS_chirp_q':
type_set = 'BNS'
if not chirp_q:
q = m2/m1
mchirp = np.power((m1*m2), 3/5) / np.power((m1+m2), 1/5)
eta = m1*m2/( (m1+m2)*(m1+m2) )
if chirp_q:
print('running chirp_q')
m1 = np.random.normal(m1, .05, 1)
m2 = np.random.normal(m2, .05, 1)
#m1 = np.ones(100) * m1
#m2 = np.ones(100) * m2
q = m2
mchirp = m1
eta = lightcurve_utils.q2eta(q)
m1, m2 = lightcurve_utils.mc2ms(mchirp, eta)
#chi_eff = np.random.uniform(-1,1,100)
chi_eff = np.ones(1)*chi
Xlan_val = 1e-3
Xlan = Xlan_val
#if lan_override:
# Xlan_val = lan_override_val
#Xlans = np.ones(1)*Xlan_val
c1 = np.ones(1)
c2 = np.ones(1)
mb1 = np.ones(1)
mb2 = np.ones(1)
data = np.vstack((m1,m2,chi_eff,mchirp,eta,q)).T
samples = KNTable((data), names = ('m1','m2','chi_eff','mchirp','eta','q'))
#data = np.vstack((m1s,m2s,dists,lambda1s,lambda2s,chi_effs,Xlans,c1,c2,mb1,mb2,mchirps,etas,qs,mej,vej, dyn_mej, wind_mej, mbnss)).T
#samples = KNTable((data), names = ('m1','m2','dist','lambda1','lambda2','chi_eff','Xlan','c1','c2','mb1','mb2','mchirp','eta','q','mej','vej', 'dyn_mej', 'wind_mej', 'mbns'))
lambda1s=[]
lambda2s=[]
m1s=[]
m2s=[]
#dists=[]
chi_effs=[]
Xlans=[]
qs=[]
etas=[]
mchirps=[]
mbnss=[]
print('...running')
nsamples = num_samples
m1s, m2s, dists_mbta = [], [], []
lambda1s, lambda2s, chi_effs = [], [], []
mbnss = []
if EOS == "gp":
# read Phil + Reed's EOS files
eospostdat = np.genfromtxt("/home/philippe.landry/nseos/eos_post_PSRs+GW170817+J0030.csv",names=True,dtype=None,delimiter=",")
idxs = np.array(eospostdat["eos"])
weights = np.array([np.exp(weight) for weight in eospostdat["logweight_total"]])
elif EOS == "Sly":
eosname = "SLy"
eos = EOS4ParameterPiecewisePolytrope(eosname)
Xlan_min, Xlan_max = -9, -1
for ii, row in enumerate(samples):
#m1, m2, dist_mbta, chi_eff = row["m1"], row["m2"], row["dist_mbta"], row["chi_eff"]
m1, m2, chi_eff = row["m1"], row["m2"], row["chi_eff"]
if EOS == "spec":
indices = np.random.randint(0, 2396, size=nsamples)
elif EOS == "gp":
indices = np.random.choice(np.arange(0,len(idxs)), size=nsamples,replace=True,p=weights/np.sum(weights))
for jj in range(nsamples):
if (EOS == "spec") or (EOS == "gp"):
index = indices[jj]
lambda1, lambda2 = -1, -1
mbns = -1
# samples lambda's from Phil + Reed's files
if EOS == "spec":
while (lambda1 < 0.) or (lambda2 < 0.) or (mbns < 0.):
eospath = "/home/philippe.landry/nseos/eos/spec/macro/macro-spec_%dcr.csv" % index
data_out = np.genfromtxt(eospath, names=True, delimiter=",")
marray, larray = data_out["M"], data_out["Lambda"]
f = interp.interp1d(marray, larray, fill_value=0, bounds_error=False)
if float(f(m1)) > lambda1: lambda1 = f(m1)
if float(f(m2)) > lambda2: lambda2 = f(m2)
if np.max(marray) > mbns: mbns = np.max(marray)
if (lambda1 < 0.) or (lambda2 < 0.) or (mbns < 0.):
index = int(np.random.randint(0, 2396, size=1)) # pick a different EOS if it returns negative Lambda or Mmax
lambda1, lambda2 = -1, -1
mbns = -1
elif EOS == "gp":
while (lambda1 < 0.) or (lambda2 < 0.) or (mbns < 0.):
phasetr = 0
eospath = "/home/philippe.landry/nseos/eos/gp/mrgagn/DRAWmod1000-%06d/MACROdraw-%06d/MACROdraw-%06d-%d.csv" % (idxs[index]/1000, idxs[index], idxs[index], phasetr)
while os.path.isfile(eospath):
data_out = np.genfromtxt(eospath, names=True, delimiter=",")
marray, larray = data_out["M"], data_out["Lambda"]
f = interp.interp1d(marray, larray, fill_value=0, bounds_error=False)
if float(f(m1)) > lambda1: lambda1 = f(m1) # pick lambda from least compact stable branch
if float(f(m2)) > lambda2: lambda2 = f(m2)
if np.max(marray) > mbns: mbns = np.max(marray) # get global maximum mass
phasetr += 1 # check all stable branches
eospath = "/home/philippe.landry/nseos/eos/gp/mrgagn/DRAWmod1000-%06d/MACROdraw-%06d/MACROdraw-%06d-%d.csv" % (idxs[index]/1000, idxs[index], idxs[index], phasetr)
if (lambda1 < 0.) or (lambda2 < 0.) or (mbns < 0.):
index = int(np.random.choice(np.arange(0,len(idxs)), size=1,replace=True,p=weights/np.sum(weights))) # pick a different EOS if it returns negative Lambda or Mmax
lambda1, lambda2 = -1, -1
mbns = -1
elif EOS == "Sly":
lambda1, lambda2 = eos.lambdaofm(m1), eos.lambdaofm(m2)
mbns = eos.maxmass()
m1s.append(m1)
m2s.append(m2)
#dists_mbta.append(dist_mbta)
lambda1s.append(lambda1)
lambda2s.append(lambda2)
chi_effs.append(chi_eff)
#Xlans.append(10**np.random.uniform(Xlan_min, Xlan_max))
Xlans.append(Xlan)
mbnss.append(mbns)
#print(len(thetas))
#check theta implementation for reproducibility
#thetas = 180. * np.arccos(np.random.uniform(-1., 1., len(samples) * nsamples)) / np.pi
idx_thetas = np.where(thetas > 90.)[0]
thetas[idx_thetas] = 180. - thetas[idx_thetas]
#thetas = list(thetas)
Xlans = np.ones(np.array(m1s).shape) * Xlan_val
#print(mbnss)
#thetas = np.ones(100)
# make final arrays of masses, distances, lambdas, spins, and lanthanide fractions
data = np.vstack((m1s,m2s,lambda1s,lambda2s,Xlans,chi_effs,thetas,mbnss)).T
samples = KNTable(data, names=('m1', 'm2', 'lambda1', 'lambda2','Xlan','chi_eff','theta', 'mbns'))
# limit masses
#samples = samples.mass_cut(mass1=3.0,mass2=3.0)
print("m1: %.5f +-%.5f"%(np.mean(samples["m1"]),np.std(samples["m1"])))
print("m2: %.5f +-%.5f"%(np.mean(samples["m2"]),np.std(samples["m2"])))
# Downsample
#samples = samples.downsample(Nsamples=100)
print(data)
print(samples)
samples = samples.calc_tidal_lambda(remove_negative_lambda=True)
print(samples)
print(samples['lambda1'], samples['lambda2'])
# Calc compactness
samples = samples.calc_compactness(fit=True)
# Calc baryonic mass
print(samples)
samples = samples.calc_baryonic_mass(EOS=None, TOV=None, fit=True)
#----------------------------------------------------------------------------------
if (not 'mej' in samples.colnames) and (not 'vej' in samples.colnames):
#mbns = 2.1
#idx1 = np.where((samples['m1'] < mbns) & (samples['m2'] < mbns))[0]
#idx2 = np.where((samples['m1'] > mbns) | (samples['m2'] > mbns))[0]
#1 BNS, 2 NSBH, 3 BBH
idx1 = np.where((samples['m1'] <= samples['mbns']) & (samples['m2'] <= samples['mbns']))[0]
idx2 = np.where((samples['m1'] > samples['mbns']) & (samples['m2'] <= samples['mbns']))[0]
idx3 = np.where((samples['m1'] > samples['mbns']) & (samples['m2'] > samples['mbns']))[0]
mej, vej = np.zeros(samples['m1'].shape), np.zeros(samples['m1'].shape)
wind_mej, dyn_mej = np.zeros(samples['m1'].shape), np.zeros(samples['m1'].shape)
#from gwemlightcurves.EjectaFits.CoDi2019 import calc_meje, calc_vej
from gwemlightcurves.EjectaFits.PaDi2019 import calc_meje, calc_vej
# calc the mass of ejecta
mej1, dyn_mej1, wind_mej1 = calc_meje(samples['m1'], samples['c1'], samples['m2'], samples['c2'], split_mej=True)
# calc the velocity of ejecta
vej1 = calc_vej(samples['m1'],samples['c1'],samples['m2'],samples['c2'])
samples['mchirp'], samples['eta'], samples['q'] = lightcurve_utils.ms2mc(samples['m1'], samples['m2'])
#samples['q'] = 1.0 / samples['q']
from gwemlightcurves.EjectaFits.KrFo2019 import calc_meje, calc_vave
# calc the mass of ejecta
mej2, dyn_mej2, wind_mej2 = calc_meje(samples['q'],samples['chi_eff'],samples['c2'], samples['m2'], split_mej=True)
# calc the velocity of ejecta
vej2 = calc_vave(samples['q'])
# calc the mass of ejecta
mej3 = np.zeros(samples['m1'].shape)
dyn_mej3 = np.zeros(samples['m1'].shape)
wind_mej3 = np.zeros(samples['m1'].shape)
# calc the velocity of ejecta
vej3 = np.zeros(samples['m1'].shape) + 0.2
mej[idx1], vej[idx1] = mej1[idx1], vej1[idx1]
mej[idx2], vej[idx2] = mej2[idx2], vej2[idx2]
mej[idx3], vej[idx3] = mej3[idx3], vej3[idx3]
wind_mej[idx1], dyn_mej[idx1] = wind_mej1[idx1], dyn_mej1[idx1]
wind_mej[idx2], dyn_mej[idx2] = wind_mej2[idx2], dyn_mej2[idx2]
wind_mej[idx3], dyn_mej[idx3] = wind_mej3[idx3], dyn_mej3[idx3]
samples['mej'] = mej
samples['vej'] = vej
samples['dyn_mej'] = dyn_mej
samples['wind_mej'] = wind_mej
# Add draw from a gaussian in the log of ejecta mass with 1-sigma size of 70%
erroropt = 'none'
if erroropt == 'none':
print("Not applying an error to mass ejecta")
elif erroropt == 'log':
samples['mej'] = np.power(10.,np.random.normal(np.log10(samples['mej']),0.236))
elif erroropt == 'lin':
samples['mej'] = np.random.normal(samples['mej'],0.72*samples['mej'])
elif erroropt == 'loggauss':
samples['mej'] = np.power(10.,np.random.normal(np.log10(samples['mej']),0.312))
#idx = np.where(samples['mej'] > 0)[0]
#samples = samples[idx]
idx = np.where(samples['mej'] <= 0)[0]
samples['mej'][idx] = 1e-11
if (model_set == "Bu2019inc"):
idx = np.where(samples['mej'] <= 1e-6)[0]
samples['mej'][idx] = 1e-11
elif (model_set == "Ka2017"):
idx = np.where(samples['mej'] <= 1e-3)[0]
samples['mej'][idx] = 1e-11
print("Probability of having ejecta")
print(100 * (len(samples) - len(idx)) /len(samples))
return samples
samples['c2'])
# calc the velocity of ejecta
vej1 = calc_vej(samples['m1'], samples['c1'], samples['m2'],
samples['c2'])
samples['mchirp'], samples['eta'], samples[
'q'] = lightcurve_utils.ms2mc(samples['m1'], samples['m2'])
samples['q'] = 1.0 / samples['q']
from gwemlightcurves.EjectaFits.KrFo2019 import calc_meje, calc_vave
# calc the mass of ejecta
mej2 = calc_meje(samples['q'], samples['chi_eff'], samples['c1'],
samples['m2'])
# calc the velocity of ejecta
vej2 = calc_vave(samples['q'])
mej[idx1], vej[idx1] = mej1[idx1], vej1[idx1]
mej[idx2], vej[idx2] = mej2[idx2], vej2[idx2]
samples['mej'] = mej
samples['vej'] = vej
# Add draw from a gaussian in the log of ejecta mass with 1-sigma size of 70%
erroropt = 'none'
if erroropt == 'none':
print("Not applying an error to mass ejecta")
elif erroropt == 'log':
samples['mej'] = np.power(
10., np.random.normal(np.log10(samples['mej']), 0.236))
elif erroropt == 'lin':
def Test(EOS,
m1,
m2,
chi,
type_set=Type,
model_set='Bu2019inc',
twixie=twixie_tf,
lan_override=False,
lan_override_val=None,
chirp_q=False):
if type_set == 'BNS_chirp_q':
type_set = 'BNS'
if not chirp_q:
m1 = np.random.normal(m1, .05, 100)
m2 = np.random.normal(m2, .05, 100)
q = m2 / m1
mchirp = np.power((m1 * m2), 3 / 5) / np.power((m1 + m2), 1 / 5)
eta = m1 * m2 / ((m1 + m2) * (m1 + m2))
if chirp_q:
print('running chirp_q')
m1 = np.random.normal(m1, .05, 100)
m2 = np.random.normal(m2, .05, 100)
#m1 = np.ones(100) * m1
#m2 = np.ones(100) * m2
q = m2
mchirp = m1
eta = lightcurve_utils.q2eta(q)
m1, m2 = lightcurve_utils.mc2ms(mchirp, eta)
dist = np.ones(100)
#chi_eff = np.random.uniform(-1,1,100)
chi_eff = np.ones(100) * chi
Xlan = 1e-3 * np.ones(100)
if lan_override:
Xlan = lan_override_val * np.ones(100)
c1 = np.ones(1000)
c2 = np.ones(1000)
mb1 = np.ones(1000)
mb2 = np.ones(1000)
mej = np.ones(1000)
vej = np.ones(1000)
data = np.vstack((m1, m2, dist, chi_eff, Xlan, mchirp, eta, q)).T
samples_tmp = KNTable(
(data),
names=('m1', 'm2', 'dist', 'chi_eff', 'Xlan', 'mchirp', 'eta', 'q'))
if twixie:
samples_tmp = KNTable.read_mchirp_samples(opts.mchirp_samples,
Nsamples=100,
twixie_flag=twixie_tf)
lambda1s = []
lambda2s = []
m1s = []
m2s = []
dists = []
chi_effs = []
Xlans = []
qs = []
etas = []
mchirps = []
mbnss = []
term1_list, term2_list, term3_list, term4_list = [], [], [], []
if EOS == "gp":
# read Phil + Reed's EOS files
filenames = glob.glob(
"/home/philippe.landry/gw170817eos/gp/macro/MACROdraw-*-0.csv")
idxs = []
for filename in filenames:
filenameSplit = filename.replace(".csv",
"").split("/")[-1].split("-")
idxs.append(int(filenameSplit[1]))
idxs = np.array(idxs)
elif EOS == "Sly":
eosname = "SLy"
eos = EOS4ParameterPiecewisePolytrope(eosname)
for ii, row in enumerate(samples_tmp):
if not twixie:
m1, m2, dist, chi_eff, q, mchirp, eta, Xlan = row["m1"], row[
"m2"], row["dist"], row["chi_eff"], row['q'], row[
'mchirp'], row['eta'], row['Xlan']
if twixie:
m1, m2, dist, chi_eff, q, mchirp, eta = row["m1"], row["m2"], row[
"dist_mbta"], row["chi_eff"], row['q'], row['mchirp'], row[
'eta']
nsamples = 10
if EOS == "spec":
indices = np.random.randint(0, 2395, size=nsamples)
elif EOS == "gp":
indices = np.random.randint(0, len(idxs), size=nsamples)
for jj in range(nsamples):
if (EOS == "spec") or (EOS == "gp"):
index = indices[jj]
# samples lambda's from Phil + Reed's files
if EOS == "spec":
eospath = "/home/philippe.landry/gw170817eos/spec/macro/macro-spec_%dcr.csv" % index
data_out = np.genfromtxt(eospath, names=True, delimiter=",")
marray, larray = data_out["M"], data_out["Lambda"]
f = interp.interp1d(marray,
larray,
fill_value=0,
bounds_error=False)
lambda1, lambda2 = f(m1), f(m2)
mbns = np.max(marray)
elif EOS == "Sly":
lambda1, lambda2 = eos.lambdaofm(m1), eos.lambdaofm(m2)
mbns = eos.maxmass()
#print(mbns)
elif EOS == "gp":
lambda1, lambda2 = 0.0, 0.0
phasetr = 0
while (lambda1 == 0.0) or (lambda2 == 0.0):
eospath = "/home/philippe.landry/gw170817eos/gp/macro/MACROdraw-%06d-%d.csv" % (
idxs[index], phasetr)
if not os.path.isfile(eospath):
break
data_out = np.genfromtxt(eospath,
names=True,
delimiter=",")
marray, larray = data_out["M"], data_out["Lambda"]
f = interp.interp1d(marray,
larray,
fill_value=0,
bounds_error=False)
lambda1_tmp, lambda2_tmp = f(m1), f(m2)
if (lambda1_tmp > 0) and (lambda1 == 0.0):
lambda1 = lambda1_tmp
if (lambda2_tmp > 0) and (lambda2 == 0.0):
lambda2 = lambda2_tmp
phasetr = phasetr + 1
mbns = np.max(marray)
lambda1s.append(lambda1)
lambda2s.append(lambda2)
m1s.append(m1)
m2s.append(m2)
dists.append(dist)
chi_effs.append(chi_eff)
Xlans.append(Xlan)
qs.append(q)
etas.append(eta)
mchirps.append(mchirp)
mbnss.append(mbns)
if twixie:
Xlans = np.ones(1000) * 1e-3
data = np.vstack((m1s, m2s, dists, lambda1s, lambda2s, chi_effs, Xlans, c1,
c2, mb1, mb2, mchirps, etas, qs, mej, vej, mbnss)).T
samples = KNTable((data),
names=('m1', 'm2', 'dist', 'lambda1', 'lambda2',
'chi_eff', 'Xlan', 'c1', 'c2', 'mb1', 'mb2',
'mchirp', 'eta', 'q', 'mej', 'vej', 'mbns'))
#calc compactness
samples = samples.calc_compactness(fit=True)
#clac baryonic mass
samples = samples.calc_baryonic_mass(EOS=None, TOV=None, fit=True)
if type_set == 'BNS':
from gwemlightcurves.EjectaFits.CoDi2019 import calc_meje, calc_vej
#from gwemlightcurves.EjectaFits.PaDi2019 import calc_meje, calc_vej
# calc the mass of ejecta
mej = calc_meje(samples['m1'], samples['c1'], samples['m2'],
samples['c2'])
# calc the velocity of ejecta
vej = calc_vej(samples['m1'], samples['c1'], samples['m2'],
samples['c2'])
samples['mchirp'], samples['eta'], samples[
'q'] = lightcurve_utils.ms2mc(samples['m1'], samples['m2'])
samples['q'] = 1.0 / samples['q']
samples['mej'] = mej
samples['vej'] = vej
if model_set == 'Bu2019inc':
idx = np.where(samples['mej'] <= 1e-6)[0]
samples['mej'][idx] = 1e-6
idx2 = np.where(samples['mej'] >= 1)[0]
samples['mej'][idx2] = 1e-6
print('mej = ' + str(samples['mej'][0]))
if type_set == 'NSBH':
from gwemlightcurves.EjectaFits.KrFo2019 import calc_meje, calc_vave
# calc the mass of ejecta
mej = calc_meje(samples['q'], samples['chi_eff'], samples['c2'],
samples['m2'])
# calc the velocity of ejecta
vej = calc_vave(samples['q'])
samples['mej'] = mej
samples['vej'] = vej
if model_set == 'Bu2019inc':
idx = np.where(samples['mej'] <= 1e-6)[0]
samples['mej'][idx] = 1e-6
idx2 = np.where(samples['mej'] >= 1)[0]
samples['mej'][idx2] = 1e-6
print('mej = ' + str(samples['mej'][0]))
if type_set == 'BNS' or 'NSBH':
# Add draw from a gaussian in the log of ejecta mass with 1-sigma size of 70%
erroropt = 'none'
if erroropt == 'none':
print("Not applying an error to mass ejecta")
elif erroropt == 'log':
samples['mej'] = np.power(
10., np.random.normal(np.log10(samples['mej']), 0.236))
elif erroropt == 'lin':
samples['mej'] = np.random.normal(samples['mej'],
0.72 * samples['mej'])
elif erroropt == 'loggauss':
samples['mej'] = np.power(
10., np.random.normal(np.log10(samples['mej']), 0.312))
idx = np.where(samples['mej'] > 0)[0]
samples = samples[idx]
print(EOS + ' calculation finished')
return samples
def calc_ejecta(self, model_KN):
idx1 = np.where((self.samples['m1'] <= self.samples['mbns'])
& (self.samples['m2'] <= self.samples['mbns']))[0]
idx2 = np.where((self.samples['m1'] > self.samples['mbns'])
& (self.samples['m2'] <= self.samples['mbns']))[0]
idx3 = np.where((self.samples['m1'] > self.samples['mbns'])
& (self.samples['m2'] > self.samples['mbns']))[0]
mej, vej = np.zeros(self.samples['m1'].shape), np.zeros(
self.samples['m1'].shape)
from gwemlightcurves.EjectaFits.CoDi2019 import calc_meje, calc_vej
# calc the mass of ejecta
mej1 = calc_meje(self.samples['m1'], self.samples['c1'],
self.samples['m2'], self.samples['c2'])
# calc the velocity of ejecta
vej1 = calc_vej(self.samples['m1'], self.samples['c1'],
self.samples['m2'], self.samples['c2'])
self.samples['mchirp'], self.samples['eta'], self.samples[
'q'] = lightcurve_utils.ms2mc(self.samples['m1'],
self.samples['m2'])
self.samples['q'] = 1.0 / self.samples['q']
from gwemlightcurves.EjectaFits.KrFo2019 import calc_meje, calc_vave
# calc the mass of ejecta
mej2 = calc_meje(self.samples['q'], self.samples['chi_eff'],
self.samples['c2'], self.samples['m2'])
# calc the velocity of ejecta
vej2 = calc_vave(self.samples['q'])
# calc the mass of ejecta
mej3 = np.zeros(self.samples['m1'].shape)
# calc the velocity of ejecta
vej3 = np.zeros(self.samples['m1'].shape) + 0.2
mej[idx1], vej[idx1] = mej1[idx1], vej1[idx1]
mej[idx2], vej[idx2] = mej2[idx2], vej2[idx2]
mej[idx3], vej[idx3] = mej3[idx3], vej3[idx3]
print(
"(mej[1284], self.samples[1284]['m1'], self.samples[1284]['m2'], self.samples[1284]['c1'], self.samples[1284]['c2'], self.samples[1284]['q'], self.samples[1284]['chi_eff'])"
)
print((mej[1284], self.samples[1284]['m1'], self.samples[1284]['m2'],
self.samples[1284]['c1'], self.samples[1284]['c2'],
self.samples[1284]['q'], self.samples[1284]['chi_eff']))
self.samples['mej'] = mej
self.samples['vej'] = vej
# Add draw from a gaussian in the log of ejecta mass with 1-sigma size of 70%
erroropt = 'none'
if erroropt == 'none':
print("Not applying an error to mass ejecta")
elif erroropt == 'log':
self.samples['mej'] = np.power(
10., np.random.normal(np.log10(self.samples['mej']), 0.236))
elif erroropt == 'lin':
self.samples['mej'] = np.random.normal(self.samples['mej'],
0.72 * self.samples['mej'])
elif erroropt == 'loggauss':
self.samples['mej'] = np.power(
10., np.random.normal(np.log10(self.samples['mej']), 0.312))
idx = np.where(self.samples['mej'] <= 0)[0]
self.samples['mej'][idx] = 1e-11
if (model_KN == "Bu2019inc"):
idx = np.where(self.samples['mej'] <= 1e-6)[0]
self.samples['mej'][idx] = 1e-11
elif (model_KN == "Ka2017"):
idx = np.where(self.samples['mej'] <= 1e-3)[0]
self.samples['mej'][idx] = 1e-11
print("Probability of having ejecta")
print(100 * (len(self.samples) - len(idx)) / len(self.samples))
return self.samples['mej']