logccdf — SciPy v1.15.3 Manual (original) (raw)

scipy.stats.Normal.

Normal.logccdf(x, y=None, /, *, method=None)[source]#

Log of the complementary cumulative distribution function

The complementary cumulative distribution function (“CCDF”), denoted\(G(x)\) is the complement of the cumulative distribution function\(F(x)\); i.e., probability the random variable \(X\) will assume a value greater than \(x\):

\[G(x) = 1 - F(x) = P(X > x)\]

A two-argument variant of this function is:

\[G(x, y) = 1 - F(x, y) = P(X < x \quad \text{or} \quad X > y)\]

logccdf computes the logarithm of the complementary cumulative distribution function (“log-CCDF”), \(\log(G(x))\)/\(\log(G(x, y))\), but it may be numerically favorable compared to the naive implementation (computing the CDF and taking the logarithm).

logccdf accepts x for \(x\) and y for \(y\).

Parameters:

x, yarray_like

The arguments of the log-CCDF. x is required; y is optional.

method{None, ‘formula’, ‘logexp’, ‘complement’, ‘quadrature’, ‘addition’}

The strategy used to evaluate the log-CCDF. By default (None), the one-argument form of the function chooses between the following options, listed in order of precedence.

• 'formula': use a formula for the log CCDF itself
• 'logexp': evaluate the CCDF and take the logarithm
• 'complement': evaluate the log-CDF and take the logarithmic complement (see Notes)
• 'quadrature': numerically log-integrate the log-PDF

The two-argument form chooses between:

• 'formula': use a formula for the log CCDF itself
• 'addition': compute the log-CDF at x and the log-CCDF at y, then take the logarithmic sum (see Notes)

Not all method options are available for all distributions. If the selected method is not available, a NotImplementedErrorwill be raised.

Returns:

outarray

The log-CCDF evaluated at the provided argument(s).

Notes

Suppose a continuous probability distribution has support \([l, r]\). The log-CCDF returns its minimum value of \(\log(0)=-\infty\) for\(x ≥ r\) and its maximum value of \(\log(1) = 0\) for\(x ≤ l\).

For distributions with infinite support, it is common forccdf to return a value of 0 when the argument is theoretically within the support; this can occur because the true value of the CCDF is too small to be represented by the chosen dtype. The log of the CCDF, however, will often be finite (not -inf) over a much larger domain. Similarly, logccdf may provided a strictly negative result with arguments for which ccdf would return 1.0. Consequently, it may be preferred to work with the logarithms of probabilities to avoid underflow and related limitations of floating point numbers.

The “logarithmic complement” of a number \(z\) is mathematically equivalent to \(\log(1-\exp(z))\), but it is computed to avoid loss of precision when \(\exp(z)\) is nearly \(0\) or \(1\). Similarly, the term “logarithmic sum” of \(w\) and \(z\)is used here to mean the \(\log(\exp(w)+\exp(z))\), AKA\(\text{LogSumExp}(w, z)\).

References

Examples

Instantiate a distribution with the desired parameters:

import numpy as np from scipy import stats X = stats.Uniform(a=-0.5, b=0.5)

Evaluate the log-CCDF at the desired argument:

X.logccdf(0.25) -1.3862943611198906 np.allclose(X.logccdf(0.), np.log(X.ccdf(0.))) True