def get_spectra_and_std(my_spec, name, test):
lb, ps = so_spectra.read_ps("spectra_%s/spectra/Dl_%s_cross.dat" %
(my_spec, name),
spectra=spectra)
lb, noise = so_spectra.read_ps("spectra_%s/spectra/Dl_%s_noise.dat" %
(my_spec, name),
spectra=spectra)
cov = np.load("covariances_%s/covariances/analytic_cov_%s_%s.npy" %
(my_spec, name, name))
cov = so_cov.selectblock(cov, ["TT", "TE", "ET", "EE"],
n_bins=len(lb),
block="%s%s" % (test, test))
std = np.sqrt(cov.diagonal())
return lb, ps[test], noise[test], cov, std
arrays_2 = d["arrays_%s" % sv2]
for id_ar2, ar2 in enumerate(arrays_2):
freq2 = d["nu_eff_%s_%s" % (sv2, ar2)]
if (id_sv1 == id_sv2) & (id_ar1 > id_ar2): continue
if (id_sv1 > id_sv2): continue
combin = "%s_%sx%s_%s" % (sv1, ar1, sv2, ar2)
spec_name = "%s_%s_%s" % (type, combin, kind)
lb, Db = so_spectra.read_ps("%s/%s.dat" % (specDir, spec_name),
spectra=spectra)
cov = np.load("%s/analytic_cov_%s_%s.npy" %
(cov_dir, combin, combin))
cov = so_cov.selectblock(cov, ["TT", "TE", "ET", "EE"],
n_bins=len(lb),
block="TT")
std = np.sqrt(cov.diagonal())
lp, clp, errorp = np.loadtxt(planck_data_dir +
"spectrum_TT_100x100.dat")
plt.errorbar(lb,
Db[spec],
std,
fmt=".",
label="%s_%s" % (spec, combin),
color=c)
# TT+TE+EE chi2
ps1_ttteee = np.concatenate([ps1[mode] for mode in modes])
ps2_ttteee = np.concatenate([ps2[mode] for mode in modes])
residual_ps = ps1_ttteee - ps2_ttteee
chi2 = residual_ps @ np.linalg.solve(res_cov, residual_ps)
chi2_dict["all", surv1, surv2, spec1,
spec2] = [chi2, len(residual_ps)]
line.append(chi2)
ndof_line.append(len(residual_ps))
# Loop on TT, TE, ET, EE
for mode in modes:
# Select block
res_covblock = so_cov.selectblock(res_cov,
modes,
n_bins=len(lb),
block=mode + mode)
# Compute power spectrum and errors
res_std = np.sqrt(res_covblock.diagonal())
res_ps = ps1[mode] - ps2[mode]
# Compute chi2
chi2 = res_ps @ np.linalg.solve(res_covblock, res_ps)
chi2_dict[mode, surv1, surv2, spec1, spec2] = [chi2, len(lb)]
line.append(chi2)
ndof_line.append(len(lb))
# Low-ell (ell < 1000) chi2
id_low = np.where(lb < 1000)
low_ps = res_ps[id_low]
for id_sv1, sv1 in enumerate(surveys):
for id_sv2, sv2 in enumerate(surveys):
if (id_sv1 > id_sv2): continue
combin = "%s_%sx%s_%s" % (sv1, ar, sv2, ar)
print("producing plots for : ", combin)
spec_name = "%s_%s_%s" % (type, combin, "cross")
lb, Db = so_spectra.read_ps("%s/%s.dat" % (specDir, spec_name),
spectra=spectra)
cov = np.load("%s/analytic_cov_%s_%s.npy" %
(cov_dir, combin, combin))
cov_select = so_cov.selectblock(cov, ["TT", "TE", "ET", "EE"],
n_bins=len(lb),
block=spec + spec)
std = np.sqrt(cov_select.diagonal())
if (sv1 == sv2):
if spec == "TE":
Db["TE"] = (Db["TE"] + Db["ET"]) / 2
_, f = ar.split("_")
f_choi = f
if f == "f220": f_choi = "f150"
if spec == "TT":
plt.semilogy()
binning_file, lmax)
n_bins = len(bin_hi)
plt.figure(figsize=(15, 15))
plt.suptitle(
"%s %s (press c/v to switch between covariance matrix elements)" %
(spec1, spec2),
fontsize=30)
count = 1
for bl in [
"TTTT", "TETE", "ETET", "EEEE", "TTTE", "TTEE", "TTET", "TEET",
"TEEE", "ETEE", "EETE", "EEET", "ETTE", "ETTT", "EETT", "TETT"
]:
mc_cov_sub = so_cov.selectblock(mc_cov, ["TT", "TE", "ET", "EE"],
n_bins,
block=bl)
analytic_cov_sub = so_cov.selectblock(analytic_cov,
["TT", "TE", "ET", "EE"],
n_bins,
block=bl)
var = mc_cov_sub.diagonal()
analytic_var = analytic_cov_sub.diagonal()
plt.subplot(3, 6, count)
if count == 1:
plt.semilogy()
plt.plot(lb[1:],
var[1:],
".",
pspy_utils.create_directory(plot_dir)
# Check effect of weighting
for bl in ["TTTT", "TETE", "EEEE"]:
for scan in scan_list:
for run in runs:
_name = "%s_%sx%s_%s" % (scan, "split0", "split0", run)
lb, _ = so_spectra.read_ps("%s/spectra_%s.dat" %
(spectra_dir, _name),
spectra=spectra)
n_bins = len(lb)
cov = np.load("%s/analytic_cov_%s_%s.npy" % (cov_dir, scan, run))
cov_select = so_cov.selectblock(cov, ["TT", "TE", "ET", "EE"],
n_bins,
block=bl)
var = cov_select.diagonal()
plt.plot(lb, np.sqrt(var), label=run)
plt.title(r"$\sigma^{%s}_{\ell}$, %s" % (bl, scan), fontsize=14)
plt.xlabel(r"$\ell$", fontsize=14)
plt.legend()
plt.savefig("%s/sigma_%s_%s.png" % (plot_dir, bl, scan),
bbox_inches="tight")
plt.clf()
plt.close()
cov_dir +
"analytic_cov_{0}_{1}.npy".format(cross_name, actxact_name))
TT_actxact = ps_actxact["TT"]
TT_actxplanck = ps_actxplanck["TT"]
id = np.where((lb >= lmin[act_map]) & (lb <= lmax))
TT_actxact = TT_actxact[id]
TT_actxplanck = TT_actxplanck[id]
TT = {"actxact": TT_actxact, "actxplanck": TT_actxplanck}
for key in cov:
cov[key] = so_cov.selectblock(cov[key],
modes,
n_bins=len(lb),
block="TTTT")
cov[key] = cov[key][np.ix_(id[0], id[0])]
spectra_and_covariances[act_map] = (TT, cov)
for act_map in reference_maps:
TT, TT_cov = spectra_and_covariances[act_map]
def loglike(c):
#PlanckxPlanck - ACTxPLanck
res_ps = c**2 * TT["actxact"] - c * TT["actxplanck"]
res_cov = c**2 * (c**2 * TT_cov["actxact_actxact"] +
TT_cov["actxplanck_actxplanck"] -
2 * c * TT_cov["actxact_actxplanck"])