plotastrodata 1.7.12__tar.gz → 1.7.14__tar.gz

{plotastrodata-1.7.12/plotastrodata.egg-info → plotastrodata-1.7.14}/PKG-INFO

@@ -1,6 +1,6 @@
 Metadata-Version: 2.4
 Name: plotastrodata
-Version: 1.7.12
+Version: 1.7.14
 Summary: plotastrodata is a tool for astronomers to create figures from FITS files and perform fundamental data analyses with ease.
 Home-page: https://github.com/yusukeaso-astron/plotastrodata
 Download-URL: https://github.com/yusukeaso-astron/plotastrodata

{plotastrodata-1.7.12 → plotastrodata-1.7.14}/plotastrodata/__init__.py

@@ -1,4 +1,4 @@
 import warnings
 
 warnings.simplefilter('ignore', FutureWarning)
-__version__ = '1.7.12'
+__version__ = '1.7.14'

{plotastrodata-1.7.12 → plotastrodata-1.7.14}/plotastrodata/fitting_utils.py

@@ -38,7 +38,7 @@ class EmceeCorner():
                  model: object | None = None,
                  xdata: np.ndarray | None = None,
                  ydata: np.ndarray | None = None,
-                 sigma: np.ndarray = 1, progressbar: bool = True,
+                 sigma: np.ndarray = 1, progressbar: bool = False,
                  percent: list = [16, 84]):
         """Make bounds, logl, and logp for ptemcee.
 
@@ -59,9 +59,12 @@ class EmceeCorner():
         else:
             global_bounds = np.array(bounds)
         global_progressbar = progressbar
-        if logl is None and None not in [model, xdata, ydata]:
+        if logl is None and (model is not None
+                             and xdata is not None
+                             and ydata is not None):
             def logl(x: np.ndarray) -> float:
-                return np.sum((ydata - model(xdata, *x))**2 / sigma**2) / (-2)
+                chi2 = np.sum((ydata - model(xdata, *x))**2 / sigma**2)
+                return chi2 / (-2)
         self.bounds = global_bounds
         self.dim = len(self.bounds)
         self.logl = logl
@@ -69,10 +72,12 @@ class EmceeCorner():
         self.percent = percent
         self.ndata = 10000 if xdata is None else len(xdata)
 
-    def fit(self, nwalkersperdim: int = 2, ntemps: int = 1, nsteps: int = 1000,
-            nburnin: int = 500, ntry: int = 1, pos0: np.ndarray | None = None,
-            savechain: str | None = None, ncores: int = 1, grcheck: bool = False,
-            pt: bool = False) -> None:
+    def fit(self, nwalkersperdim: int = 2,
+            ntemps: int = 1, nsteps: int = 1000,
+            nburnin: int = 500, ntry: int = 1,
+            pos0: np.ndarray | None = None,
+            savechain: str | None = None, ncores: int = 1,
+            grcheck: bool = False, pt: bool = False) -> None:
         """Perform a Markov Chain Monte Carlo (MCMC) fitting process using the ptemcee library, which is a parallel tempering version of the emcee package, and make a corner plot of the samples using the corner package.
 
         Args:

{plotastrodata-1.7.12 → plotastrodata-1.7.14}/plotastrodata/other_utils.py

@@ -1,9 +1,9 @@
 import warnings
 import numpy as np
-from scipy.optimize import curve_fit
 from scipy.special import erf
 from scipy.interpolate import RegularGridInterpolator as RGI
 
+from plotastrodata.fitting_utils import EmceeCorner
 from plotastrodata.matrix_utils import Mrot, dot2d
 
 
@@ -37,102 +37,124 @@ def isdeg(s: str) -> bool:
         return False
 
 
+def _estimate_rms_hist(data: np.ndarray, sigma: str) -> tuple:
+    h_range = (-3.5, 3.5)
+    h = np.linspace(*h_range, 101)
+    dh = 0.07
+    m0 = np.mean(data)
+    s0 = np.std(data)
+    hist, hbin = np.histogram((data - m0) / s0, bins=100,
+                              density=True, range=h_range)
+    hbin = (hbin[:-1] + hbin[1:]) / 2
+    
+    def normalize(f):
+        """Decorator to normalize a function over h_range."""
+        def wrapper(x, *args):
+            area = np.sum(f(h, *args)) * dh
+            if area == 0:
+                p = np.where(np.abs(x - args[1]) < dh / 2, 1 / dh, 0)
+            else:
+                p = f(x, *args) / area
+            return p
+        return wrapper
+
+    if 'pbcor' in sigma:
+        @normalize
+        def model(x, *args):
+            s, m, R = args
+            x1 = (x - m) / np.sqrt(2) / s
+            x0 = (x * 2**(-R**2) - m) / np.sqrt(2) / s
+            p = erf(x1) - erf(x0)
+            p = p / (2 * np.log(2) * x * R**2)
+            return p
+        bounds = [[0.1, 2], [-2, 2], [0.1, 2]]
+    else:
+        @normalize
+        def model(x, *args):
+            s, m = args
+            x1 = (x - m) / np.sqrt(2) / s
+            p = np.exp(-x1**2)
+            p = p / (np.sqrt(2 * np.pi) * s)
+            return p
+        bounds = [[0.1, 2], [-2, 2]]
+    # curve_fit does not work for this fitting.
+    fitter = EmceeCorner(bounds=bounds, sigma=np.max(hist) * 0.01,
+                         model=model, xdata=hbin, ydata=hist)
+    fitter.fit(nwalkersperdim=4, nsteps=200, nburnin=0)
+    popt = fitter.popt
+    ave = popt[1] * s0 + m0
+    noise = popt[0] * s0
+    return ave, noise
+
+
 def estimate_rms(data: np.ndarray, sigma: float | str | None = 'hist'
                  ) -> float:
     """Estimate a noise level of a N-D array.
        When a float number or None is given, this function just outputs it.
-       Following methos are acceptable.
+       The following methods are acceptable for data selection. Multiple options are possible.
        'edge': use data[0] and data[-1].
+       'out': exclude inner 60% about axes=-2 and -1.
        'neg': use only negative values.
-       'med': use the median of data^2 assuming Gaussian.
        'iter': exclude outliers.
-       'out': exclude inner 60% about axes=-2 and -1.
-       'hist': fit histgram with Gaussian. This option can be combined with 'edge' and/or 'neg'.
-       'hist-pbcor': fit histgram with PB-corrected Gaussian. This option can be combined with 'edge' and/or 'neg'.
+       The following methods are acceptable for noise estimation. Only single option is possible.
+       'med': calculate rms from the median of data^2 assuming Gaussian.
+       'hist': fit histgram with Gaussian.
+       'hist-pbcor': fit histgram with PB-corrected Gaussian.
+       '(no string)': calculate the mean and standard deviation.
 
     Args:
         data (np.ndarray): N-D array.
-        sigma (float or str): One of the methods above. Defaults to 'hist'.
+        sigma (float or str): Methods above, like 'edge,neg,hist-pbcor'. Defaults to 'hist'.
 
     Returns:
-        float: the estimated standard deviation of noise.
+        float: The estimated standard deviation of noise.
     """
-    if sigma is None:
-        return None
-
-    def warning_offset(ave, noise):
-        if np.abs(ave) > 0.2 * noise:
-            s = 'The intensity offset is larger than 0.2 sigma.'
-            warnings.warn(s, UserWarning)
-
     nums = [float, int, np.float64, np.int64, np.float32, np.int32]
-    if type(sigma) in nums:
-        noise = sigma
-    elif np.ndim(np.squeeze(data)) == 0:
+    if sigma is None or type(sigma) in nums:
+        return sigma
+
+    if np.ndim(np.squeeze(data)) == 0:
         print('sigma cannot be estimated from only one pixel.')
-        noise = 0.0
-    elif sigma == 'edge':
-        ave = np.nanmean(data[::len(data) - 1])
-        noise = np.nanstd(data[::len(data) - 1])
-        warning_offset(ave, noise)
-    elif sigma == 'neg':
-        noise = np.sqrt(np.nanmean(data[data < 0]**2))
-    elif sigma == 'med':
-        noise = np.sqrt(np.nanmedian(data**2) / 0.454936)
-    elif sigma == 'iter':
-        n = data.copy()
-        for _ in range(5):
-            ave, sig = np.nanmean(n), np.nanstd(n)
-            n = n - ave
-            n = n[np.abs(n) < 3.5 * sig]
-        ave = np.nanmean(n)
-        noise = np.nanstd(n)
-        warning_offset(ave, noise)
-    elif sigma == 'out':
-        n, n0, n1 = data * 1, len(data), len(data[0])
-        n = np.moveaxis(n, [-2, -1], [0, 1])
-        n[n0//5: n0*4//5, n1//5: n1*4//5] = np.nan
-        if np.all(np.isnan(n)):
-            print('sigma=\'neg\' instead of \'out\' because'
-                  + ' the outer region is filled with nan.')
-            noise = np.sqrt(np.nanmean(data[data < 0]**2))
+        return 0.0
+
+    # Selection
+    n = data * 1
+    if 'edge' in sigma:
+        if np.ndim(n) <= 2:
+            print('\'edge\' is ignored because ndim <= 2.')
         else:
-            ave = np.nanmean(n)
-            noise = np.nanstd(n)
-            warning_offset(ave, noise)
-    elif 'hist' in sigma:
-        n = data * 1
-        if 'edge' in sigma:
             n = n[::len(n) - 1]
-        if 'neg' in sigma:
-            n = n[n < 0]
-            n = np.r_[n, -n]
-        n = n[~np.isnan(n)]
-        m0, s0 = np.mean(n), np.std(n)
-        hist, hbin = np.histogram(n, bins=100, density=True,
-                                  range=(m0 - s0 * 3.5,
-                                         m0 + s0 * 3.5))
-        hist, hbin = hist * s0, (hbin[:-1] + hbin[1:]) / 2 / s0
-        if 'pbcor' in sigma:
-            def g(x, s, c, R):
-                y2 = (x - c) / np.sqrt(2) / s
-                y1 = (x * 2**(-R**2) - c) / np.sqrt(2) / s
-                p = erf(y2) - erf(y1)
-                p = p / (2 * np.log(2) * x * R**2)
-                return p
-            popt, _ = curve_fit(g, hbin, hist, p0=[1, 0, 1],
-                                bounds=[[0.001, -2, 0.001], [2, 2, 2]])
+    if 'out' in sigma and 'pbcor' in sigma:
+        print('\'out\' is ignored because of \'pbcor\'.')
+    elif 'out' in sigma:
+        nx = np.shape(n)[-1]
+        ny = np.shape(n)[-2]
+        ntmp = np.moveaxis(n, [-2, -1], [0, 1])
+        ntmp[ny // 5 : ny * 4 // 5, nx // 5 : nx * 4 // 5] = np.nan
+        if np.all(np.isnan(ntmp)):
+            print('\'out\' is ignored because'
+                  + ' the outer region is filled with nan.')
         else:
-            def g(x, s, c):
-                xn = (x - c) / np.sqrt(2) / s
-                p = np.exp(-xn**2) / np.sqrt(2 * np.pi) / s
-                return p
-            popt, _ = curve_fit(g, hbin, hist, p0=[1, 0],
-                                bounds=[[0.001, -2], [2, 2]])
-        ave = popt[1]
-        noise = popt[0]
-        warning_offset(ave, noise)
-        noise = noise * s0
+            n = ntmp
+    n = n[~np.isnan(n)]
+    if 'neg' in sigma:
+        n = n[n < 0]
+        n = np.r_[n, -n]
+    if 'iter' in sigma:
+        for _ in range(5):
+            n = n[np.abs(n - np.mean(n)) < 3.5 * np.std(n)]
+    # Estimation
+    if 'hist' in sigma:
+        ave, noise = _estimate_rms_hist(n, sigma)
+    elif 'med' in sigma:
+        ave = 0
+        noise = np.sqrt(np.median(n**2) / 0.454936)
+    else:
+        ave = np.mean(n)
+        noise = np.std(n)
+    if np.abs(ave) > 0.2 * noise:
+        s = 'The intensity offset is larger than 0.2 sigma.'
+        warnings.warn(s, UserWarning)
     return noise
 
 

{plotastrodata-1.7.12 → plotastrodata-1.7.14/plotastrodata.egg-info}/PKG-INFO

@@ -1,6 +1,6 @@
 Metadata-Version: 2.4
 Name: plotastrodata
-Version: 1.7.12
+Version: 1.7.14
 Summary: plotastrodata is a tool for astronomers to create figures from FITS files and perform fundamental data analyses with ease.
 Home-page: https://github.com/yusukeaso-astron/plotastrodata
 Download-URL: https://github.com/yusukeaso-astron/plotastrodata