add script to plot residuals with respect to the best fit model

project/data_analysis/dr6.rst

@@ -248,3 +248,4 @@ In order to do plots for the paper it is useful to apply them to the spectra (an
     OMP_NUM_THREADS=256 srun -n 1 -c 256 --cpu-bind=cores python results_plot_TT.py post_likelihood.dict
     OMP_NUM_THREADS=256 srun -n 1 -c 256 --cpu-bind=cores python results_plot_combined_spectra.py post_likelihood.dict
     OMP_NUM_THREADS=256 srun -n 1 -c 256 --cpu-bind=cores python results_plot_with_planck.py post_likelihood.dict
+    OMP_NUM_THREADS=256 srun -n 1 -c 256 --cpu-bind=cores python results_plot_residuals.py post_likelihood.dict

project/data_analysis/python/paper_plots/results_plot_residuals.py

@@ -0,0 +1,171 @@
+"""
+This script plot the residuals with respect to the LCDM model specified in the dictionnary file
+"""
+
+from pspy import so_dict, so_spectra, pspy_utils
+from pspipe_utils import  log, best_fits, external_data, covariance
+import numpy as np
+import pylab as plt
+import sys, os
+from matplotlib import rcParams
+import pspipe_utils
+import scipy.stats as ss
+
+rcParams["font.family"] = "serif"
+rcParams["font.size"] = "12"
+rcParams["xtick.labelsize"] = 12
+rcParams["ytick.labelsize"] = 12
+rcParams["axes.labelsize"] = 12
+rcParams["axes.titlesize"] = 12
+
+d = so_dict.so_dict()
+d.read_from_file(sys.argv[1])
+log = log.get_logger(**d)
+
+tag = d["best_fit_tag"]
+binning_file = d["binning_file"]
+lmax = d["lmax"]
+
+combined_spec_dir = f"combined_spectra{tag}"
+bestfit_dir = f"best_fits{tag}"
+
+run_name = {}
+run_name["_paper"] = "ACT"
+run_name["_paper_PACT"] = "PACT"
+run_name["_Planck"] = "Planck"
+run_name["_Planck_LB"] = "Planck LB"
+
+plot_dir = f"plots/combined_spectra{tag}/"
+pspy_utils.create_directory(plot_dir)
+
+type = d["type"]
+
+planck_data_path = os.path.join(os.path.dirname(os.path.abspath(pspipe_utils.__file__)), "data/spectra/planck")
+
+########################################################################################
+selected_spectra_list = [["TT"], ["EE"], ["TE", "ET"]]
+########################################################################################
+
+ylim = {}
+ylim["TT"] = [10, 7000]
+ylim["TE"] = [-105000, 75000]
+ylim["EE"] = [0, 45]
+
+ylim_res = {}
+ylim_res["TT"] =  [-120000, 120000]
+ylim_res["TE"] = [-15000, 15000]
+ylim_res["EE"] = [-4, 4]
+
+fac = {}
+fac["TT"] = 0
+fac["TE"] = 1
+fac["EE"] = 0
+res_fac = {}
+res_fac["TT"] = 1
+res_fac["TE"] = 1
+res_fac["EE"] = 0
+
+y_ticks_res = {}
+y_ticks_res["TT"] = [-100000, -50000, 0 , 50000, 100000]
+y_ticks_res["EE"] =  [-3, -2, -1, 0, 1, 2, 3]
+y_ticks_res["TE"] = [-10000, -5000, 0, 5000, 10000]
+
+
+spectra = ["TT", "TE", "TB", "ET", "BT", "EE", "EB", "BE", "BB"]
+
+lth, Dlth = so_spectra.read_ps(f"{bestfit_dir}/cmb.dat", spectra=spectra)
+Dlb_th,Dlb_th_large = {}, {}
+for spectrum in spectra:
+    lb_th, Dlb_th[spectrum] = pspy_utils.naive_binning(lth, Dlth[spectrum], binning_file, lmax)
+    lb_th_large, Dlb_th_large[spectrum] = pspy_utils.naive_binning(lth, Dlth[spectrum], "../dr6/BIN_ACTPOL_50_4_SC_large_bin_at_low_ell", lmax)
+
+
+l_planck, ps_planck_b, sigma_planck, cov_planck = external_data.get_planck_cmb_only_data()
+l_b, ps_th_plank = external_data.bin_ala_planck_cmb_only(lth, Dlth)
+
+rebin_fac = 2
+
+for spec_select in selected_spectra_list:
+    s_name = spec_select[0]
+
+    lb_ml, vec_ml, sigma_ml = np.loadtxt(f"{combined_spec_dir}/{type}_all_{s_name}_cmb_only.dat", unpack=True)
+    cov_ml = np.load(f"{combined_spec_dir}/cov_all_{s_name}.npy")
+    inv_cov_ml = np.linalg.inv(cov_ml)
+
+    l_p = l_planck[s_name]
+    f_p = l_p * (l_p + 1) / (2 * np.pi)
+    Dl_p = ps_planck_b[s_name] * f_p
+    sigma_p = sigma_planck[s_name] * f_p
+    cov_p = cov_planck[s_name+s_name] * np.outer(f_p, f_p)
+    Dl_p_th = ps_th_plank[s_name] * f_p
+
+    l_p_rebin, Dl_p_rebin, cov_rebin = covariance.rebin_spectrum_with_cov(l_p, Dl_p, cov_p, rebin_fac=rebin_fac)
+    _, Dl_p_th_rebin, _ = covariance.rebin_spectrum_with_cov(l_p, Dl_p_th, cov_p, rebin_fac=rebin_fac)
+
+
+    id = np.where(lb_th >= lb_ml[0])
+    res = (vec_ml -  Dlb_th[s_name][id])
+    chi2 = res @ inv_cov_ml @ res
+    ndof = len(lb_ml)
+    pte = 1 - ss.chi2(ndof).cdf(chi2)
+
+
+    res_p = (Dl_p - Dl_p_th)
+    chi_2_p = res_p @ np.linalg.inv(cov_p) @ res_p
+    ndof_p = len(Dl_p)
+    pte_p = 1 - ss.chi2(ndof_p).cdf(chi_2_p)
+
+    res_p_rebin = (Dl_p_rebin - Dl_p_th_rebin)
+    chi_2_p_rebin = res_p_rebin @ np.linalg.inv(cov_rebin) @ res_p_rebin
+    ndof_p_rebin = len(Dl_p_rebin)
+    pte_p_rebin = 1 - ss.chi2(ndof_p_rebin).cdf(chi_2_p_rebin)
+    sigma_p_rebin = np.sqrt(cov_rebin.diagonal())
+
+
+    print(pte, pte_p, pte_p_rebin)
+
+    plt.figure(figsize=(16, 8))
+    plt.subplot(2,1,1)
+    if s_name == "TT": plt.semilogy()
+    plt.ylim(ylim[s_name])
+    plt.errorbar(lb_ml, vec_ml *  lb_ml ** fac[s_name], sigma_ml * lb_ml ** fac[s_name] , fmt=".", color="royalblue", label="ACT")
+    plt.errorbar(l_p_rebin, Dl_p_rebin * l_p_rebin ** fac[s_name], sigma_p_rebin * l_p_rebin ** fac[s_name], fmt=".", color="darkorange", alpha=0.6, label="Planck")
+    plt.plot(lth, Dlth[s_name] * lth ** fac[s_name], color="gray", alpha=1, label=r" %s $\Lambda$CDM" % run_name[tag])
+
+    if fac[s_name] == 0:
+        plt.ylabel(r"$D^{%s}_{\ell}$" % s_name, fontsize=22)
+    if fac[s_name] == 1:
+        plt.ylabel(r"$\ell D^{%s}_{\ell}$" % s_name, fontsize=22)
+    if fac[s_name] > 1:
+        plt.ylabel(r"$\ell^{%s}D^{%s}_{\ell}$" % (fac[s_name], s_name), fontsize=22)
+
+    plt.xlim(0,4000)
+    plt.legend(fontsize=16)
+    plt.xticks([])
+
+    plt.subplot(2,1,2)
+    plt.xlabel(r"$\ell$", fontsize=22)
+
+    if res_fac[s_name] == 0:
+        plt.ylabel(r"$(D^{%s}_{\ell} - D^{%s, th}_{\ell, \rm %s}) $" % ( s_name, s_name, run_name[tag]), fontsize=22)
+    if res_fac[s_name] == 1:
+        plt.ylabel(r"$\ell (D^{%s}_{\ell} - D^{%s, th}_{\ell, \rm %s}) $" % (s_name, s_name, run_name[tag]), fontsize=22)
+    if res_fac[s_name] > 1:
+        plt.ylabel(r"$\ell^{%s} (D^{%s}_{\ell} - D^{%s, th}_{\ell, \rm %s}) $" % (res_fac[s_name], s_name, s_name, run_name[tag]), fontsize=22)
+
+
+    plt.errorbar(lb_ml, res  *  lb_ml ** res_fac[s_name], sigma_ml  *  lb_ml ** res_fac[s_name],
+                 label=f"ACT (PTE: {pte:.3f})", fmt=".", color="royalblue")
+    plt.errorbar(l_p_rebin, res_p_rebin  *  l_p_rebin ** res_fac[s_name], sigma_p_rebin  *  l_p_rebin ** res_fac[s_name],
+                 label=f"Planck (PTE: {pte_p_rebin:.3f})", alpha=0.6, fmt=".", color="darkorange")
+
+    plt.plot(lb_th, lb_th * 0, color="gray")
+    plt.xlim(0,4000)
+    plt.ylim(ylim_res[s_name])
+    plt.yticks(ticks=y_ticks_res[s_name], labels=y_ticks_res[s_name])
+    plt.legend(fontsize=16)
+    plt.subplots_adjust(wspace=0, hspace=0)
+    plt.savefig(f"{plot_dir}/residal_vs_best_fit_cmb_{s_name}.png", bbox_inches="tight")
+    plt.clf()
+    plt.close()
+