winsorizing, finally without the extra D

doc/index.rst

@@ -30,8 +30,6 @@ citing MSNoise:
 for Monitoring Seismic Velocity Changes Using Ambient Seismic Noise,
 *Seismological Research Letters*, 85(3), 715‑726, doi:10.1785/0220130073.
 
-This documentation is also available in PDF format on the MSNoise Website
-(PDF_).
 
 Installation
 ============

doc/installation.rst

@@ -12,7 +12,7 @@ Note that MSNoise is always tested against the latest release versions of the ma
 
 To run MSNoise, you need:
 
-* A recent version of Python (3.x recommended). We suggest using Miniconda_
+* A recent version of Python (3.11 recommended). We suggest using Miniconda_
   and creating a fresh environment for running msnoise.
   MSNoise is tested "continuously" on GitHub for the last 3 most recent python version, and the three OS.
 

doc/releasenotes/msnoise-1.3.rst

@@ -254,7 +254,7 @@ Some improvements to the maths have been done for MSNoise 1.3:
 * ``whiten``: the symmetric hermitian was not properly defined and could lead to
   a 1 sample shift in the negative frequencies.
 * ``compute_cc``: it is now possible to define an overlap of the windows
-* ``compute_cc``: setting ``windsorizing`` to ``-1`` now computes the 1-bit
+* ``compute_cc``: setting ``winsorizing`` to ``-1`` now computes the 1-bit
   normalization of the trace. Reminder: ``0``: no normalization, ``N``: N*rms
   clipping.
 * ``mwcs``: the tapering of the windowed CCF has been improved in order to

doc/releasenotes/msnoise-1.5.rst

@@ -319,7 +319,7 @@ Improvements in terms of performances have also been done for MSNoise 1.5:
 * Added fftpack optimized nfft (scipy's next_fast_len). This could lead to some
   small differences in the final result of the MWCS procedure, because of the
   number of points used for smoothing the (cross-)spectra.
-* Replaced binarization (sign) and windsorizing (clip) by standard numpy
+* Replaced binarization (sign) and winsorizing (clip) by standard numpy
   functions operating directly inplace on the arrays, avoiding unecessary
   copies.
 * The preprocessing only reads files that should contain the right component.

msnoise/default.csv

@@ -22,7 +22,7 @@ response_prefilt,"(0.005, 0.006, 30.0, 35.0)","Remove instrument correction pre-
 maxlag,120,Maximum lag (in seconds),float,,,
 corr_duration,1800,Data windows to correlate (in seconds),float,,,
 overlap,0,Amount of overlap between data windows [0:1[,float,,,
-windsorizing,3,"Windsorizing at N time RMS , 0 disables windsorizing, -1 enables 1-bit normalization",float,,,
+winsorizing,3,"Winsorizing at N time RMS , 0 disables winsorizing, -1 enables 1-bit normalization",float,,,
 whitening,A,"Whiten Traces before cross-correlation: [A]ll (except for autocorr), [N]one, or only if [C]omponents are different",str,A/N/C,,
 whitening_type,B,"Type of spectral whitening function to use: [B]rutal (amplitude to 1.0), divide spectrum by its [PSD] or by band-passing the white spectrum with a hanning-window [HANN]. WARNING: only works for compute_cc, not compute_cc_rot, where it will always be B",str,B/PSD/HANN,,
 clip_after_whiten,N,"Do the clipping (winsorizing) after whitening?",bool,Y/N,,

msnoise/s03compute_cc.py

@@ -14,7 +14,7 @@
 * |overlap|
 * |maxlag|
 * |corr_duration|
-* |windsorizing|
+* |winsorizing|
 * |resampling_method|
 * |remove_response|
 * |response_format|
@@ -71,7 +71,7 @@
 *station* table. The rotation is computed for Radial and Transverse components.
 
 Then, for each ``corr_duration`` window in the signal, and for each filter
-configured in the database, the traces are clipped to ``windsorizing`` times
+configured in the database, the traces are clipped to ``winsorizing`` times
 the RMS (or 1-bit converted) and then whitened in the frequency domain
 (see :ref:`whiten`) between the frequency bounds. The whitening procedure can be
 skipped by setting the ``whitening`` configuration to `None`. The two other
@@ -340,10 +340,10 @@ def main(loglevel="INFO"):
                     tmp.detrend("demean")
 
                     for tr in tmp:
-                        if params.windsorizing == -1:
+                        if params.winsorizing == -1:
                             np.sign(tr.data, tr.data)  # inplace
-                        elif params.windsorizing != 0:
-                            rms = tr.data.std() * params.windsorizing
+                        elif params.winsorizing != 0:
+                            rms = tr.data.std() * params.winsorizing
                             np.clip(tr.data, -rms, rms, tr.data)  # inplace
 
                     # TODO should not hardcode 4 percent!

msnoise/s03compute_no_rotation.py

@@ -51,7 +51,7 @@
 * |overlap|
 * |maxlag|
 * |corr_duration|
-* |windsorizing|
+* |winsorizing|
 * |resampling_method|
 * |remove_response|
 * |response_format|
@@ -102,7 +102,7 @@
 *station* table. The rotation is computed for Radial and Transverse components.
 
 Then, for each ``corr_duration`` window in the signal, and for each filter
-configured in the database, the traces are clipped to ``windsorizing`` times
+configured in the database, the traces are clipped to ``winsorizing`` times
 the RMS (or 1-bit converted) and then whitened in the frequency domain
 (see :ref:`whiten`) between the frequency bounds. The whitening procedure can be
 skipped by setting the ``whitening`` configuration to `None`. The two other
@@ -227,15 +227,15 @@ def winsorizing(data, params, input="timeseries", nfft=0):
         data = sf.ifftn(data, [nfft, ], axes=[1, ]).astype(float)
 
     for i in range(data.shape[0]):
-        if params.windsorizing == -1:
+        if params.winsorizing == -1:
             np.sign(data[i], data[i])  # inplace
-        elif params.windsorizing != 0:
+        elif params.winsorizing != 0:
 
             # imin, imax = scoreatpercentile(data[i], [0.1, 99.9])
             # not_outliers = np.where((data[i] >= imin) &
             #                         (data[i] <= imax))[0]
-            # rms = data[i][not_outliers].std() * params.windsorizing
-            rms = data[i].std() * params.windsorizing
+            # rms = data[i][not_outliers].std() * params.winsorizing
+            rms = data[i].std() * params.winsorizing
             np.clip(data[i], -rms, rms, data[i])  # inplace
     if input == "fft":
         data = sf.fftn(data, [nfft, ], axes=[1, ])