import copy
import numpy
from matplotlib import pyplot
from .probscale import _minimal_norm
from . import validate
from . import algo
[docs]def probplot(data, ax=None, plottype='prob', dist=None, probax='x',
problabel=None, datascale='linear', datalabel=None,
bestfit=False, return_best_fit_results=False,
estimate_ci=False, ci_kws=None, pp_kws=None,
scatter_kws=None, line_kws=None, **fgkwargs):
"""
Probability, percentile, and quantile plots.
Parameters
----------
data : array-like
1-dimensional data to be plotted
ax : matplotlib axes, optional
The Axes on which to plot. If one is not provided, a new Axes
will be created.
plottype : string (default = 'prob')
Type of plot to be created. Options are:
- 'prob': probabilty plot
- 'pp': percentile plot
- 'qq': quantile plot
dist : scipy distribution, optional
A distribtion to compute the scale's tick positions. If not
specified, a standard normal distribution will be used.
probax : string, optional (default = 'x')
The axis ('x' or 'y') that will serve as the probability (or
quantile) axis.
problabel, datalabel : string, optional
Axis labels for the probability/quantile and data axes
respectively.
datascale : string, optional (default = 'log')
Scale for the other axis that is not
bestfit : bool, optional (default is False)
Specifies whether a best-fit line should be added to the plot.
return_best_fit_results : bool (default is False)
If True a dictionary of results of is returned along with the
figure.
estimate_ci : bool, optional (False)
Estimate and draw a confidence band around the best-fit line
using a percentile bootstrap.
ci_kws : dict, optional
Dictionary of keyword arguments passed directly to
``viz.fit_line`` when computing the best-fit line.
pp_kws : dict, optional
Dictionary of keyword arguments passed directly to
``viz.plot_pos`` when computing the plotting positions.
scatter_kws, line_kws : dict, optional
Dictionary of keyword arguments passed directly to ``ax.plot``
when drawing the scatter points and best-fit line, respectively.
Other Parameters
----------------
color : string, optional
A directly-specified matplotlib color argument for both the
data series and the best-fit line if drawn. This argument is
made available for compatibility for the seaborn package and
is not recommended for general use. Instead colors should be
specified within ``scatter_kws`` and ``line_kws``.
.. note::
Users should not specify this parameter. It is inteded to
only be used by seaborn when operating within a
``FacetGrid``.
label : string, optional
A directly-specified legend label for the data series. This
argument is made available for compatibility for the seaborn
package and is not recommended for general use. Instead the
data series label should be specified within ``scatter_kws``.
.. note::
Users should not specify this parameter. It is inteded to
only be used by seaborn when operating within a
``FacetGrid``.
Returns
-------
fig : matplotlib.Figure
The figure on which the plot was drawn.
result : dict of linear fit results, optional
Keys are:
- q : array of quantiles
- x, y : arrays of data passed to function
- xhat, yhat : arrays of modeled data plotted in best-fit line
- res : array of coeffcients of the best-fit line.
See also
--------
viz.plot_pos
viz.fit_line
numpy.polyfit
scipy.stats.probplot
scipy.stats.mstats.plotting_positions
Examples
--------
Probability plot with the probabilities on the y-axis
.. plot::
:context: close-figs
>>> import numpy; numpy.random.seed(0)
>>> from matplotlib import pyplot
>>> from scipy import stats
>>> from probscale.viz import probplot
>>> data = numpy.random.normal(loc=5, scale=1.25, size=37)
>>> fig = probplot(data, plottype='prob', probax='y',
... problabel='Non-exceedance probability',
... datalabel='Observed values', bestfit=True,
... line_kws=dict(linestyle='--', linewidth=2),
... scatter_kws=dict(marker='o', alpha=0.5))
Quantile plot with the quantiles on the x-axis
.. plot::
:context: close-figs
>>> fig = probplot(data, plottype='qq', probax='x',
... problabel='Theoretical Quantiles',
... datalabel='Observed values', bestfit=True,
... line_kws=dict(linestyle='-', linewidth=2),
... scatter_kws=dict(marker='s', alpha=0.5))
"""
if dist is None:
dist = _minimal_norm
# check input values
fig, ax = validate.axes_object(ax)
probax = validate.axis_name(probax, 'probability axis')
problabel = validate.axis_label(problabel)
datalabel = validate.axis_label(datalabel)
# default values for symbology options
scatter_kws = validate.other_options(scatter_kws)
line_kws = validate.other_options(line_kws)
pp_kws = validate.other_options(pp_kws)
# check plottype
plottype = validate.axis_type(plottype)
## !-- kwarg that only seaborn should use --! ##
_color = fgkwargs.get('color', None)
if _color is not None:
scatter_kws['color'] = _color
line_kws['color'] = _color
## !-- kwarg that only seaborn should use --! ##
_label = fgkwargs.get('label', None)
if _label is not None:
scatter_kws['label'] = _label
# compute the plotting positions and sort the data
probs, datavals = plot_pos(data, **pp_kws)
qntls = dist.ppf(probs)
# determine how the probability values should be expressed
if plottype == 'qq':
probvals = qntls
else:
probvals = probs * 100
# set up x, y, Axes for probabilities on the x
if probax == 'x':
x, y = probvals, datavals
ax.set_xlabel(problabel)
ax.set_ylabel(datalabel)
if plottype == 'prob':
ax.set_xscale('prob', dist=dist)
fitprobs = 'x'
else:
fitprobs = None
if plottype == 'pp':
ax.set_xlim(left=0, right=100)
ax.set_yscale(datascale)
fitlogs = 'y' if datascale == 'log' else None
# setup x, y, Axes for probabilities on the y
elif probax == 'y':
y, x = probvals, datavals
ax.set_xlabel(datalabel)
ax.set_ylabel(problabel)
if plottype == 'prob':
ax.set_yscale('prob', dist=dist)
fitprobs = 'y'
else:
fitprobs = None
if plottype == 'pp':
ax.set_ylim(bottom=0, top=100)
ax.set_xscale(datascale)
fitlogs = 'x' if datascale == 'log' else None
# finally plot the data
linestyle = scatter_kws.pop('linestyle', 'none')
marker = scatter_kws.pop('marker', 'o')
ax.plot(x, y, linestyle=linestyle, marker=marker, **scatter_kws)
# maybe do a best-fit and plot
if bestfit:
xhat, yhat, modelres = fit_line(x, y, xhat=sorted(x), dist=dist,
fitprobs=fitprobs, fitlogs=fitlogs,
estimate_ci=estimate_ci)
ax.plot(xhat, yhat, **line_kws)
if estimate_ci:
# for alpha, use half of existing or 0.5 * 0.5 = 0.25
# for zorder, use 1 less than existing or 1 - 1 = 0
opts = {
'facecolor': line_kws.get('color', 'k'),
'edgecolor': line_kws.get('color', 'None'),
'alpha': line_kws.get('alpha', 0.5) * 0.5,
'zorder': line_kws.get('zorder', 1) - 1,
'label': '95% conf. interval'
}
ax.fill_between(xhat, y1=modelres['yhat_hi'], y2=modelres['yhat_lo'],
**opts)
else:
xhat, yhat, modelres = (None, None, None)
# set the probability axes limits
if plottype == 'prob':
_set_prob_limits(ax, probax, len(probs))
# return the figure and maybe results of the best-fit
if return_best_fit_results:
results = dict(q=qntls, x=x, y=y, xhat=xhat, yhat=yhat, res=modelres)
return fig, results
else:
return fig
[docs]def plot_pos(data, postype=None, alpha=None, beta=None):
"""
Compute the plotting positions for a dataset. Heavily borrows from
``scipy.stats.mstats.plotting_positions``.
A plottiting position is defined as: ``(i-alpha)/(n+1-alpha-beta)``
where:
- ``i`` is the rank order
- ``n`` is the size of the dataset
- ``alpha`` and ``beta`` are parameters used to adjust the
positions.
The values of ``alpha`` and ``beta`` can be explicitly set. Typical
values can also be access via the ``postype`` parameter. Available
``postype`` values (alpha, beta) are:
"type 4" (alpha=0, beta=1)
Linear interpolation of the empirical CDF.
"type 5" or "hazen" (alpha=0.5, beta=0.5)
Piecewise linear interpolation.
"type 6" or "weibull" (alpha=0, beta=0)
Weibull plotting positions. Unbiased exceedance probability
for all distributions. Recommended for hydrologic
applications.
"type 7" (alpha=1, beta=1)
The default values in R. Not recommended with probability
scales as the min and max data points get plotting positions
of 0 and 1, respectively, and therefore cannot be shown.
"type 8" (alpha=1/3, beta=1/3)
Approximately median-unbiased.
"type 9" or "blom" (alpha=0.375, beta=0.375)
Approximately unbiased positions if the data are normally
distributed.
"median" (alpha=0.3175, beta=0.3175)
Median exceedance probabilities for all distributions
(used in ``scipy.stats.probplot``).
"apl" or "pwm" (alpha=0.35, beta=0.35)
Used with probability-weighted moments.
"cunnane" (alpha=0.4, beta=0.4)
Nearly unbiased quantiles for normally distributed data.
This is the default value.
"gringorten" (alpha=0.44, beta=0.44)
Used for Gumble distributions.
Parameters
----------
data : array-like
The values whose plotting positions need to be computed.
postype : string, optional (default: "cunnane")
alpha, beta : float, optional
Custom plotting position parameters is the options available
through the `postype` parameter are insufficient.
Returns
-------
plot_pos : numpy.array
The computed plotting positions, sorted.
data_sorted : numpy.array
The original data values, sorted.
References
----------
http://artax.karlin.mff.cuni.cz/r-help/library/lmomco/html/pp.html
http://astrostatistics.psu.edu/su07/R/html/stats/html/quantile.html
https://scipy-docs.dev.org.tw/doc/scipy-0.17.0/reference/generated/scipy.stats.probplot.html
https://scipy-docs.dev.org.tw/doc/scipy-0.17.0/reference/generated/scipy.stats.mstats.plotting_positions.html
"""
pos_params = {
'type 4': (0, 1),
'type 5': (0.5, 0.5),
'type 6': (0, 0),
'type 7': (1, 1),
'type 8': (1/3., 1/3.),
'type 9': (0.375, 0.375),
'weibull': (0, 0),
'median': (0.3175, 0.3175),
'apl': (0.35, 0.35),
'pwm': (0.35, 0.35),
'blom': (0.375, 0.375),
'hazen': (0.5, 0.5),
'cunnane': (0.4, 0.4),
'gringorten': (0.44, 0.44), # Gumble
}
postype = 'cunnane' if postype is None else postype
if alpha is None and beta is None:
alpha, beta = pos_params[postype.lower()]
data = numpy.asarray(data, dtype=float).flatten()
n = data.shape[0]
pos = numpy.empty_like(data)
pos[n:] = 0
sorted_index = data.argsort()
pos[sorted_index[:n]] = (numpy.arange(1, n+1) - alpha) / (n + 1.0 - alpha - beta)
return pos[sorted_index], data[sorted_index]
def _set_prob_limits(ax, probax, N):
""" Sets the limits of a probabilty axis based the number of point.
Parameters
----------
ax : matplotlib Axes
The Axes object that will be modified.
N : int
Maximum number of points for the series plotted on the Axes.
which : string
The axis whose ticklabels will be rotated. Valid values are 'x',
'y', or 'both'.
Returns
-------
None
"""
fig, ax = validate.axes_object(ax)
which = validate.axis_name(probax, 'probability axis')
if N <= 5:
minval = 10
elif N <= 10:
minval = 5
else:
minval = 10 ** (-1 * numpy.ceil(numpy.log10(N) - 2))
if which in ['x', 'both']:
ax.set_xlim(left=minval, right=100-minval)
elif which in ['y', 'both']:
ax.set_ylim(bottom=minval, top=100-minval)
[docs]def fit_line(x, y, xhat=None, fitprobs=None, fitlogs=None, dist=None,
estimate_ci=False, niter=10000, alpha=0.05):
"""
Fits a line to x-y data in various forms (linear, log, prob scales).
Parameters
----------
x, y : array-like
Independent and dependent data, respectively.
xhat : array-like, optional
The values at which ``yhat`` should should be estimated. If
not provided, falls back to the sorted values of ``x``.
fitprobs, fitlogs : str, optional.
Defines how data should be transformed. Valid values are
'x', 'y', or 'both'. If using ``fitprobs``, variables should
be expressed as a percentage, i.e.,
for a probablility transform, data will be transformed with
``lambda x: dist.ppf(x / 100.)``.
For a log transform, ``lambda x: numpy.log(x)``.
Take care to not pass the same value to both ``fitlogs`` and
``figprobs`` as both transforms will be applied.
dist : distribution, optional
A fully-spec'd scipy.stats distribution-like object
such that ``dist.ppf`` and ``dist.cdf`` can be called. If not
provided, defaults to a minimal implementation of
``scipt.stats.norm``.
estimate_ci : bool, optional (False)
Estimate and draw a confidence band around the best-fit line
using a percentile bootstrap.
niter : int, optional (default = 10000)
Number of bootstrap iterations if ``estimate_ci`` is provided.
alpha : float, optional (default = 0.05)
The confidence level of the bootstrap estimate.
Returns
-------
xhat, yhat : numpy arrays
Linear model estimates of ``x`` and ``y``.
results : dict
Dictionary of linear fit results. Keys include:
- slope
- intersept
- yhat_lo (lower confidence interval of the estimated y-vals)
- yhat_hi (upper confidence interval of the estimated y-vals)
"""
fitprobs = validate.fit_argument(fitprobs, "fitprobs")
fitlogs = validate.fit_argument(fitlogs, "fitlogs")
# maybe set xhat to default values
if xhat is None:
xhat = copy.copy(x)
# maybe set dist to default value
if dist is None:
dist = _minimal_norm
# maybe compute ppf of x
if fitprobs in ['x', 'both']:
x = dist.ppf(x / 100.)
xhat = dist.ppf(numpy.array(xhat)/100.)
# maybe compute ppf of y
if fitprobs in ['y', 'both']:
y = dist.ppf(y / 100.)
# maybe compute log of x
if fitlogs in ['x', 'both']:
x = numpy.log(x)
# maybe compute log of y
if fitlogs in ['y', 'both']:
y = numpy.log(y)
yhat, results = algo._fit_simple(x, y, xhat, fitlogs=fitlogs)
if estimate_ci:
yhat_lo, yhat_hi = algo._bs_fit(x, y, xhat, fitlogs=fitlogs,
niter=niter, alpha=alpha)
else:
yhat_lo, yhat_hi = None, None
# maybe undo the ppf transform
if fitprobs in ['y', 'both']:
yhat = 100. * dist.cdf(yhat)
if yhat_lo is not None:
yhat_lo = 100. * dist.cdf(yhat_lo)
yhat_hi = 100. * dist.cdf(yhat_hi)
# maybe undo ppf transform
if fitprobs in ['x', 'both']:
xhat = 100. * dist.cdf(xhat)
results['yhat_lo'] = yhat_lo
results['yhat_hi'] = yhat_hi
return xhat, yhat, results