[frames] | no frames]

# Module Numeric_gpk

source code

 Classes EdgesTooWide
Functions

 add_overlap(a, astart, b, bstart) Add arrays a and b in the overlap region. source code

 Poisson(nbar) Return a Poisson random integer, whose distribution has a mean = nbar. source code

 bevel_concat(a, b, bevel=0, bevel_overlap=1.0, delay=0, ta=None, tb=None) Concatenate two time series. source code

 argmax(a) source code

 N_maximum(a) source code

 N_minimum(a) source code

 N_frac_rank(a, fr) source code

 N_mean_ad(a) Mean absolute deviation. source code

 median_across(list_of_vec) Returns an element-by-element median of a list of Numeric vectors. source code

 N_median(a, axis=0) Returns an element-by-element median of a list of Numeric vectors. source code

 N_median_across(a) source code

 set_diag(x, a) Set the diagonal of a matrix x to be the vector a. source code

 limit(low, x, high) source code

 trimmed_mean_sigma_across(list_of_vec, weights, clip) source code

 trimmed_mean_across(list_of_vec, weights, clip) source code

 trimmed_stdev_across(list_of_vec, weights, clip) source code

 vec_variance(x) Take a component-by-component variance of a list of vectors. source code

 qform(vec, mat) A quadratic form: vec*mat*vec, or vecs*mat*transpose(vecs) source code

 KolmogorovSmirnov(d1, d2, w1=None, w2=None) source code

 interpN(a, t) Interpolate array a to floating point indices, t, via nearest-neighbor interpolation. source code
numpy array.
 interp(a, t) Interpolate to a specified time axis. source code

 split_into_clumps(x, threshold, minsize=1) This reports when the signal is above the threshold. source code

 block_stdev(x) This is just a alternative implementation of the standard deviation of each channel, but it is designed in a block-wise fashion so the total memory usage is not large. source code

 convolve(x, kernel) This is basically like numpy.convolve(x, kernel, 1) except that it properly handles the case where the kernel is longer than the data. source code

 edge_window(n, eleft, eright, typeleft='linear', typeright='linear', norm=None) Creates a window which is basically flat, but tapers off on the left and right edges. source code

 edge_window_t(t, eleft, eright, typeleft='linear', typeright='linear', norm=None) Computes a window which is basically flat, but tapers off on the left and right edges. source code
 Variables pylab = None hash(x) asinh = acosh = BLOCK = 8192 __package__ = 'gmisclib'

Imports: numpy, math, variance, stdev, make_diag

 Function Details

source code

Add arrays a and b in the overlap region. Return (data, start). If a, b are time series, they are assumed to have the same sampling rate. Astart and Bstart apply to the zeroth index. All other indices are assumed to match start and length.

### bevel_concat(a, b, bevel=0, bevel_overlap=1.0, delay=0, ta=None, tb=None)

source code

Concatenate two time series. Bevel the edges, and overlap them slightly.

Bevel_overlap controls the fractional overlap of the two bevels, and delay specifies an extra delay for b.

If ta and/or tb are specified, return a tuple of (concatenated_time_series, tma, tmb) where tma and tmb are the locations corresponding to ta and tb in the corresponding input arrays.

source code

Mean absolute deviation. For a multi-dimensional array, it takes the MAD along the first axis, so N_mean_ad(x)[0]==N_mean_ad(x[:,0]).

### set_diag(x, a)

source code

Set the diagonal of a matrix x to be the vector a. If a is shorter than the diagonal of x, just set the beginning.

### interpN(a, t)

source code

Interpolate array a to floating point indices, t, via nearest-neighbor interpolation. Returns a Numeric array.

### interp(a, t)

source code

Interpolate to a specified time axis. This does a linear interpolation. A is a Numpy array, and t is an array of times.

Returns: numpy array.
interpolated values

### split_into_clumps(x, threshold, minsize=1)

source code

This reports when the signal is above the threshold.

Parameters:
• x (numpy.ndarray, one-dimensional.) - a signal
• threshold (float) - a threshold.
Returns:
[(start, stop), ...] for each region ("clump") where x>threshold.

### edge_window(n, eleft, eright, typeleft='linear', typeright='linear', norm=None)

source code

Creates a window which is basically flat, but tapers off on the left and right edges. The widths of the tapers can be controlled, as can the shapes.

Parameters:
• n (int) - the width of the window
• eleft (int) - the width of the left taper (i.e. at zero index, in samples).
• eright (int) - the width of the right taper (i.e. at index near n, in samples).
• typeleft ('linear' or 'cos') - what kind of taper on the left? (Defaults to "linear").
• typeright ('linear' or 'cos') - what kind of taper on the right? (Defaults to "linear").
• norm (None or float != 0.) - How to normalize? The default is None, which means no normalization. Providing a number x will normalize the window so that the average of the window**x==1.

### edge_window_t(t, eleft, eright, typeleft='linear', typeright='linear', norm=None)

source code

Computes a window which is basically flat, but tapers off on the left and right edges. The widths of the tapers can be controlled, as can the shapes.

Parameters:
• t (numpy.ndarray) - an array of time values. They are required to be monotonically increasing, and assumed to be linearly spaced.
• eleft (float) - the width of the left taper (i.e. at zero index, in time units).
• eright (float) - the width of the right taper (i.e. at index near n, in time units).
• typeleft ('linear' or 'cos') - what kind of taper on the left? (Defaults to "linear").
• typeright ('linear' or 'cos') - what kind of taper on the right? (Defaults to "linear").
• norm (None or float != 0.) - How to normalize? The default is None, which means no normalization. Providing a number x will normalize the window so that the average of the window**x==1.

 Generated by Epydoc 3.0.1 on Thu Sep 22 04:25:02 2011 http://epydoc.sourceforge.net