Simple average shifted histograms (ASHs) in python (1D and 2D).
After installation, the module can be imported using
import ash
Given some values (vals), a number of bins (nbins), and number of shifts (nshifts), you can calculate a 1D ASH using
bins, heights = ash.ash1d(vals, nbins, nshifts)
Similarly, given values along two dimensions, (xvals) and (yvals), you can calculate a 2D ASH using
binsGrid, heights2d = ash.ash2d(xvals, yvals, nbins, nshifts)
By default, the samples are assumed to be weighted equally. If not, weights for each sample can be passed using the optional weights argument. If the data are periodic (such as values of an angle from 0 to 2π), periodic boundary conditions can be specified by setting periodic=True.
The upper and lower limits for the histograms can be specified using the range argument to ash1d (e.g. range=(xmin, xmax) or the xrange and yrange arguments to ash2d. Otherwise, the minimum and maximum data values will be used.
Once computed, ASHs are straightforward to plot, e.g.
import matplotlib.pyplot as plt
plt.plot(bins, heights)
...
plt.pcolormesh(binsGrid[0], binsGrid[1], heights2d)
Sampling from a mixture of three Gaussian distributions:
Sampling from a mixture of two 2D Gaussian distributions:
First, clone the repository using,
git clone https://github.com/ajdittmann/ash.git.
Then, install with
pip install -e .
or
python setup.py install
.