def plot_bf(i, args, bfpath, savepath=None):
cosmo, h, Omega_m = get_cosmo(i)
M = args['M']
params = np.loadtxt(bfpath)
colors = [
plt.get_cmap("seismic")(ci) for ci in np.linspace(1.0, 0.0, len(zs))
]
fig, ax = plt.subplots(2, sharex=True)
for j in range(len(zs)):
Mlo, Mhi, N, Mave = AD.building_box_binned_mass_function(i, j).T
edge = 10**np.concatenate((Mlo, Mhi[-1:]))
cov = AD.building_box_binned_mass_function_covariance(i, j)
err = np.sqrt(cov.diagonal())
s2 = args['sigma2s'][j]
s2t = args['sigma2tops'][j]
s2b = args['sigma2bots'][j]
d, e, f, g = model_swap(params, args, x[j])
dndM = massfunction._dndM_sigma2_precomputed(M, s2, s2t, s2b, Omega_m,
d, e, f, g)
Nmodel = massfunction.n_in_bins(edge, M, dndM) * volume
chi2 = np.dot(N - Nmodel, np.dot(np.linalg.inv(cov), N - Nmodel))
print "Box%d snap%d chi2 = %.2f" % (i, j, chi2)
ax[0].errorbar(Mave, N, err, c=colors[j], marker='.', ls='')
ax[0].loglog(Mave, Nmodel, ls='-', c=colors[j])
pd = (Nmodel - N) / Nmodel
pde = err / Nmodel
ax[1].errorbar(Mave, pd, pde, c=colors[j])
ax[0].set_ylim(1, 1e6)
yl = .11
ax[1].set_ylim(-yl, yl)
plt.subplots_adjust(hspace=0)
if savepath:
plt.gcf().savefig(savepath)
else:
plt.show()
def get_box_resids(i):
cosmo, h, Omega_m = get_cosmo(i)
M = np.logspace(12, 16, num=200) #Msun/h
k = np.logspace(-5, 1, num=1000) #Mpc^-1
bfpath = "bf_%s_box%d.txt" % (name, i)
print i, name
params = np.loadtxt(bfpath)
zb = np.array([])
Mb = np.array([])
nub = np.array([])
R = np.array([])
Re = np.array([])
for j in range(0, 10): #snap
Mlo, Mhi, N, Mtot = AD.building_box_binned_mass_function(i, j).T
Mj = Mtot / N
Mb = np.concatenate((Mb, Mj))
zb = np.concatenate((zb, np.ones_like(N) * zs[j]))
edge = 10**np.concatenate((Mlo, Mhi[-1:]))
cov = AD.building_box_binned_mass_function_covariance(i, j)
icov = np.linalg.inv(cov)
err = np.sqrt(np.diag(cov))
p = np.array([cosmo.pk_lin(ki, zs[j]) for ki in k]) * h**3
nub = np.concatenate((nub, bias.nu_at_M(Mj, k / h, p, Omega_m)))
d, e, f, g = model_swap(params, name, x[j])
dndM = massfunction.dndM_at_M(M, k / h, p, Omega_m, d, e, f, g)
Nmodel = massfunction.n_in_bins(edge, M, dndM) * volume
R = np.concatenate((R, (Nmodel - N) / Nmodel))
Re = np.concatenate((Re, err / Nmodel))
return Mb, nub, zb, R, Re
def test_mf_binned():
M2 = np.logspace(12, 16, num=100)
dn = mf.dndM_at_M(M2, k / h, p, Omega_m)
edges = np.logspace(12, 16, 11)
n = mf.n_in_bins(edges, M2, dn)
npt.assert_array_less(n[1:], n[:-1])
n2 = np.array([
mf.n_in_bin(edges[i], edges[i + 1], M2, dn)
for i in range(len(edges) - 1)
])
npt.assert_array_less(n2[1:], n2[:-1])
npt.assert_array_equal(n, n2)
def mass_func(pars, args):
x = args['x']
Omega_m = args['Omega_m']
M = args['M']
k = args['k']
ps = args['ps']
edges = args['edges'] #Msun/h
Nout = []
for i in range(len(x)):
d, e, f, g = model_swap(pars, args['name'], x[i])
dndM = massfunction.dndM_at_M(M, k, ps[i], Omega_m, d, e, f, g)
Nout.append(massfunction.n_in_bins(edges[i], M, dndM) * volume)
return Nout
def lnlike(params, args):
x = args['x'] # a - 0.5
Om = args['Omega_m']
M = args['M']
s2s = args['sigma2s']
s2ts = args['sigma2tops']
s2bs = args['sigma2bots']
Ns = args['Ns'] #Number of halos in the sim
edges = args['edges'] #Msun/h
icovs = args['icovs']
LL = 0
for i in range(len(x)):
d, e, f, g = model_swap(params, args, x[i])
dndM = massfunction._dndM_sigma2_precomputed(M, s2s[i], s2ts[i],
s2bs[i], Om, d, e, f, g)
N = massfunction.n_in_bins(edges[i], M, dndM) * volume
X = Ns[i] - N
LL += np.dot(X, np.dot(icovs[i], X))
return -0.5 * LL