def __call__(self, x):
""" Returns a linearly interpolated value of the distribution at x.
Parameters
----------
x : array-like or float
x-values for which you want to know the corresponding y-values
of the distribution.
Raises
------
ValueError
If x is outside the data range.
"""
if self._x is None and isinstance(self._y, (float, np.float)):
if isinstance(x, (list, tuple, np.ndarray)):
return np.zeros(len(x)) + self._y
else:
return self._y
if not allinrange(x, self._x_range):
raise ValueError("x is outside data range.", {"x": x, "x_range": self._x_range})
if self.hist:
idx = np.searchsorted(self._edges[:-1], x)
return self._y[idx]
else:
y = np.interp(x, self._x, self._y, left=np.nan, right=np.nan)
return y
def lookup(self, x):
""" Returns the probability from the cumulative distribution function
corresponding to the value x.
Parameters
----------
x : array-like or float
The x-value must be in the data range.
Returns
-------
prob : array-like or float
The corresponding probability of the value.
Raises
------
ValueError
If any element of x is not inside the data range.
Example
-------
This method is useful is you need to truncate the distribution
before sampling. For example::
import numpy as np
import matplotlib.pyplot as plt
x = np.linspace(0, 200, 200)
y = np.exp(-((x - 50.0)/20)**2)
dist = Distribution(x, y)
p1 = dist.lookup(40.0)
p2 = dist.lookup(80.0)
draw = dist.sample(np.random.uniform(p1, p2, 10000))
plt.hist(draw)
"""
if not allinrange(x, self._x_range):
raise ValueError("x is outside data range.", {"x": x, "x_range": self._x_range})
if self.hist:
idx = np.searchsorted(self._edges[:-1], x)
return self._cdf[idx]
else:
prob = np.interp(x, self._x, self._cdf, left=np.nan, right=np.nan)
if prob.size == 1:
prob = prob.tolist() # actually a float
return prob
def sample(self, p):
""" Returns the value corresponding to the probability from the cumulative
distribution function.
Parameters
----------
p : array-like or float
The probability value, must be between 0 and 1.
Returns
-------
xval : array-like or float
The corresponding x-value.
Raises
------
ValueError
If p is outside the range [0, 1].
Example
-------
Use this method to sample the distribution with randoms numbers. For
example::
import numpy as np
import matplotlib.pyplot as plt
x = np.linspace(0, 200, 200)
y = np.exp(-((x - 50.0)/20)**2)
dist = Distribution(x, y)
drawn = dist.sample(np.random.uniform(0, 1, 10000))
plt.hist(drawn)
"""
if not allinrange(p, (0.0, 1.0)):
raise ValueError("p is outside valid range.")
if self.hist:
idx = np.searchsorted(self._cdf, p)
try:
return self._x[idx]
except IndexError:
return self._x[-1]
else:
xval = np.interp(p, self._cdf, self._x, left=np.nan, right=np.nan)
if xval.size == 1:
xval = xval.tolist() # actually a float
return xval
def __call__(self, x):
""" Returns a linearly interpolated value of the distribution at x.
Parameters
----------
x : array-like or float
x-values for which you want to know the corresponding y-values
of the distribution.
Raises
------
ValueError
If x is outside the data range.
"""
if not allinrange(x, self._x_range):
raise ValueError("x is outside data range.")
y = np.interp(x, self._x, self._y, left=np.nan, right=np.nan)
return y
def lookup(self, x):
""" Returns the probability from the cumulative distribution function
corresponding to the value x.
Parameters
----------
x : array-like or float
The x-value must be in the data range.
Returns
-------
prob : array-like or float
The corresponding probability of the value.
Raises
------
ValueError
If any element of x is not inside the data range.
Example
-------
This method is useful is you need to truncate the distribution
before sampling. For example::
import numpy as np
import matplotlib.pyplot as plt
x = np.linspace(0, 200, 200)
y = np.exp(-((x - 50.0)/20)**2)
dist = Distribution(x, y)
p1 = dist.lookup(40.0)
p2 = dist.lookup(80.0)
draw = dist.sample(np.random.uniform(p1, p2, 10000))
plt.hist(draw)
"""
if not allinrange(x, self._x_range):
raise ValueError("x is outside data range.")
prob = np.interp(x, self._x, self._cdf, left=np.nan, right=np.nan)
if prob.size == 1:
prob = prob.tolist() # actually a float
return prob
def sample(self, p):
""" Returns the value corresponding to the probability from the cumulative
distribution function.
Parameters
----------
p : array-like or float
The probability value, must be between 0 and 1.
Returns
-------
xval : array-like or float
The corresponding x-value.
Raises
------
ValueError
If p is outside the range [0, 1].
Example
-------
Use this method to sample the distribution with randoms numbers. For
example::
import numpy as np
import matplotlib.pyplot as plt
x = np.linspace(0, 200, 200)
y = np.exp(-((x - 50.0)/20)**2)
dist = Distribution(x, y)
drawn = dist.sample(np.random.uniform(0, 1, 10000))
plt.hist(drawn)
"""
if not allinrange(p, (0.0, 1.0)):
raise ValueError("p is outside valid range.")
xval = np.interp(p, self._cdf, self._x, left=np.nan, right=np.nan)
if xval.size == 1:
xval = xval.tolist() # actually a float
return xval