results = compute_best_frequencies(ids, n_eval=10000, n_retry=5)
#------------------------------------------------------------
# Plot the phased light-curves
fig = plt.figure(figsize=(5, 6.5))
fig.subplots_adjust(hspace=0.1, bottom=0.06, top=0.94, left=0.12, right=0.94)
for i in range(6):
# get the data and best-fit angular frequency
t, y, dy = data[ids[i]].T
omega, power = results[ids[i]]
omega_best = omega[np.argmax(power)]
print " - omega_0 = %.10g" % omega_best
# do a fit to the first 4 Fourier components
mtf = MultiTermFit(omega_best, 4)
mtf.fit(t, y, dy)
phase_fit, y_fit, phased_t = mtf.predict(1000, return_phased_times=True)
# plot the phased data and best-fit curves
ax = fig.add_subplot(321 + i)
ax.errorbar(phased_t, y, dy, fmt='.k', ecolor='gray',
lw=1, ms=4, capsize=1.5)
ax.plot(phase_fit, y_fit, '-b', lw=2)
ax.set_xlim(0, 1)
ax.set_ylim(plt.ylim()[::-1])
ax.yaxis.set_major_locator(plt.MaxNLocator(4))
ax.text(0.03, 0.04, "ID = %i" % ids[i], ha='left', va='bottom',
transform=ax.transAxes)
for f in factors:
for n in nterms:
PSDs[(f, n)] = multiterm_periodogram(t, y, dy, omega / f, n)
# Compute the best-fit omega from the 6-term fit
omega_best = dict()
for f in factors:
omegaf = omega / f
PSDf = PSDs[(f, 6)]
omega_best[f] = omegaf[np.argmax(PSDf)]
# Compute the best-fit solution based on the fundamental frequency
best_fit = dict()
for f in factors:
for n in nterms:
mtf = MultiTermFit(omega_best[f], n)
mtf.fit(t, y, dy)
phase_best, y_best = mtf.predict(1000, adjust_offset=False)
best_fit[(f, n)] = (phase_best, y_best)
#------------------------------------------------------------
# Plot the results
fig = plt.figure(figsize=(12, 6))
fig.subplots_adjust(left=0.07, right=0.95, wspace=0.18,
bottom=0.1, top=0.95, hspace=0.12)
for i, f in enumerate(factors):
P_best = 2 * np.pi / omega_best[f]
phase_best = (t / P_best) % 1
for f in factors:
for n in nterms:
PSDs[(f, n)] = multiterm_periodogram(t, y, dy, omega / f, n)
# Compute the best-fit omega from the 6-term fit
omega_best = dict()
for f in factors:
omegaf = omega / f
PSDf = PSDs[(f, 6)]
omega_best[f] = omegaf[np.argmax(PSDf)]
# Compute the best-fit solution based on the fundamental frequency
best_fit = dict()
for f in factors:
for n in nterms:
mtf = MultiTermFit(omega_best[f], n)
mtf.fit(t, y, dy)
phase_best, y_best = mtf.predict(1000, adjust_offset=False)
best_fit[(f, n)] = (phase_best, y_best)
#------------------------------------------------------------
# Plot the results
fig = plt.figure(figsize=(5, 2.5))
fig.subplots_adjust(left=0.1, right=0.95, wspace=0.25,
bottom=0.12, top=0.95, hspace=0.2)
for i, f in enumerate(factors):
P_best = 2 * np.pi / omega_best[f]
phase_best = (t / P_best) % 1
for i, window in enumerate(windows):
plt.figure()
## A ##
# get the data and best-fit angular frequency
tA = data['t_eval'][data['window_id']==window]
yA = data['sig_evalA'][data['window_id']==window]
yA = (yA - np.mean(yA)) / np.std(yA)
dyA = data['sig_errA'][data['window_id']==window]
dyA = (dyA - np.mean(dyA)) / np.std(dyA)
omegaA, powerA = resultsA[window]
omega_bestA = omegaA[np.argmax(powerA)]
print " - omega_0A = %.10g" % omega_bestA
# do a fit to the first 4 Fourier components
mtfA = MultiTermFit(omega_bestA, 10)
mtfA.fit(tA, yA, dyA)
phase_fitA, y_fitA, phased_tA = mtfA.predict(1000, return_phased_times=True)
plt.errorbar(phased_tA, yA, dyA,
fmt='.r', ecolor='gray',
lw=1, ms=4, capsize=1.5)
plt.plot(phase_fitA, y_fitA, '-b', lw=2, label="P_A = %.2f day" %(2 * np.pi / omega_bestA * 24. * 365.))
## B ##
# get the data and best-fit angular frequency
tB = data['t_eval'][data['window_id']==window]
yB = data['sig_evalB'][data['window_id']==window]
yB = (yB - np.mean(yB)) / np.std(yB)
dyB = data['sig_errB'][data['window_id']==window]
dyB = (dyB - np.mean(dyB)) / np.std(dyB)
omegaB, powerB = resultsB[window]
fluxByPhase = [f for _, f in sorted(zip(phase, astDiffFlux))]
errByPhase = [f for _, f in sorted(zip(phase, astFluxErr))]
sortedPhase = sorted(phase)
sortedPhaseBin = sorted(rtp.binning(phase, time, binW, ind)[0])
fluxByPhaseBin = rtp.binWeighted(fluxByPhase, time, astFluxErr, binW,
ind)[0]
weights = [1. / (err**2) for err in errByPhase]
pfit = np.poly1d(np.polyfit(sortedPhase, fluxByPhase, deg=d, w=weights))
fitCurve = pfit(sortedPhase)
strCurve = fluxByPhase / fitCurve
fluxByPhaseSG = savitzky_golay(fluxByPhase, window_size=wsSG, order=4)
#----- Fourier Fit ------------------------------------------------------------
omega = 2 * np.pi / Prot * 24.
mtf = MultiTermFit(omega, ffOrder)
mtf.fit(time, astDiffFlux, astFluxErr)
phaseFit, fluxFit, phasedTime = mtf.predict(1000, return_phased_times=True)
phaseFit -= phasedTime[0]
phasedTime -= phasedTime[0]
phasedTime = [1 + t if t < 0 else t for t in phasedTime]
phaseFit = [1 + t if t < 0 else t for t in phaseFit]
plt.errorbar(phasedTime,
astDiffFlux,
astFluxErr,
fmt='.k',
ecolor='gray',
lw=1,
ms=4,
