LogNormalFactory

(Source code, png)

../../_images/openturns-LogNormalFactory-1.png
class LogNormalFactory(*args)

Lognormal factory distribution.

Notes

Several estimators to build a LogNormal distribution from a scalar sample are proposed. The default strategy is using the local likelihood maximum estimator.

Moments based estimator:

Lets denote:

  • \displaystyle \overline{x}_n = \frac{1}{n} \sum_{i=1}^n x_i the empirical mean of the sample,

  • \displaystyle s_n^2 = \frac{1}{n-1} \sum_{i=1}^n (x_i - \overline{x}_n)^2 its empirical variance,

  • \displaystyle a_{3,n} = \sqrt{n} \frac{\sum_{i=1}^n (x_i - \overline{x}_n)^3}{ \left( \sum_{i=1}^n (x_i - \overline{x}_n)^2 \right)^{3/2}} its empirical skewness.

We note \omega = e^{\sigma_l^2}. The estimator \hat{\omega}_n of \omega is the positive root of the relation:

(1)\omega^3 + 3 \omega^2 - (4 + a_{3,n}^2) = 0

Then we estimate (\hat{\mu}_{ln}, \hat{\sigma}_{ln}, \hat{\gamma}_{n}) using:

(2)\hat{\mu}_{ln} &= \log \hat{\beta}_{n} \\
\hat{\sigma}_{ln} &= \sqrt{ \log \hat{\omega}_{n} } \\
\hat{\gamma}_{ln} &= \overline{x}_n - \hat{\beta}_{n} \sqrt{ \hat{\omega}_{n} }

where \displaystyle \hat{\beta}_{n} = \frac{s_n}{\hat{\omega}_{n} (\hat{\omega}_{n} - 1)}.

Modified moments based estimator:

Using \overline{x}_n and s_n^2 previously defined, the third equation is:

(3)\Eset[ \log (X_{(1)} - \gamma)] = \log (x_{(1)} - \gamma)

The quantity \displaystyle EZ_1 (n) = \frac{\Eset[ \log (X_{(1)} - \gamma)] - \mu_l}{\sigma_l} is the mean of the first order statistics of a standard normal sample of size n. We have:

(4)EZ_1(n) = \int_\Rset nz\phi(z) (1 - \Phi(z))^{n-1}\di{z}

where \varphi and \Phi are the PDF and CDF of the standard normal distribution. The estimator \hat{\omega}_{n} of \omega is obtained as the solution of:

(5)\omega (\omega - 1) - \kappa_n \left[ \sqrt{\omega} - e^{EZ_1(n)\sqrt{\log \omega}} \right]^2 = 0

where \displaystyle \kappa_n = \frac{s_n^2}{(\overline{x}_n - x_{(1)})^2}. Then we have (\hat{\mu}_{ln}, \hat{\sigma}_{ln}, \hat{\gamma}_{n}) using the relations defined for the moments based estimator (2).

Local maximum likelihood estimator:

The following sums are defined:

S_0 &= \sum_{i=1}^n \frac{1}{x_i - \gamma} \\
S_1 &= \sum_{i=1}^n \log (x_i - \gamma) \\
S_2 &= \sum_{i=1}^n \log^2 (x_i - \gamma) \\
S_3 &= \sum_{i=1}^n \frac{\log (x_i - \gamma)}{x_i - \gamma}

The Maximum Likelihood estimator of (\mu_{l}, \sigma_{l}, \gamma) is defined by:

(6)\hat{\mu}_{l,n} &= \frac{S_1(\hat{\gamma})}{n} \\
\hat{\sigma}_{l,n}^2 &= \frac{S_2(\hat{\gamma})}{n} - \hat{\mu}_{l,n}^2

Thus, \hat{\gamma}_n satisfies the relation:

(7)S_0 (\gamma) \left(S_2(\gamma) - S_1(\gamma) \left( 1 + \frac{S_1(\gamma)}{n} \right) \right) + n S_3(\gamma) = 0

under the constraint \gamma \leq \min x_i.

Least squares method estimator:

The parameter \gamma is numerically optimized by non-linear least-squares:

\min{\gamma} \norm{\Phi^{-1}(\hat{F}_n(x_i)) - (a_1 \log(x_i - \gamma) + a_0)}_2^2

where a_0, a_1 are computed from linear least-squares at each optimization evaluation.

When \gamma is known and the x_i follow a Log-Normal distribution then we use linear least-squares to solve the relation:

(8)\Phi^{-1}(\hat{F}_n(x_i)) = a_1 \log(x_i - \gamma) + a_0

And the remaining parameters are estimated with:

\hat{\sigma}_l &= \frac{1}{a_1}\\
\hat{\mu}_l &= -a_0 \hat{\sigma}_l

Examples

>>> import openturns as ot
>>> ot.RandomGenerator.SetSeed(0)
>>> sample = ot.LogNormal(1.5, 2.5, -1.5).getSample(1000)
>>> estimated = ot.LogNormalFactory().build(sample)

Methods

build(*args)

Build the distribution.

buildAsLogNormal(*args)

Build the distribution as a LogNormal type.

buildEstimator(*args)

Build the distribution and the parameter distribution.

buildMethodOfLeastSquares(sample)

Build the distribution based on the least-squares estimator.

buildMethodOfLocalLikelihoodMaximization(sample)

Build the distribution based on the local likelihood maximum estimator.

buildMethodOfModifiedMoments(sample)

Build the distribution based on the modified moments estimator.

buildMethodOfMoments(sample)

Build the distribution based on the method of moments estimator.

getBootstrapSize()

Accessor to the bootstrap size.

getClassName()

Accessor to the object's name.

getName()

Accessor to the object's name.

hasName()

Test if the object is named.

setBootstrapSize(bootstrapSize)

Accessor to the bootstrap size.

setName(name)

Accessor to the object's name.

__init__(*args)
build(*args)

Build the distribution.

Available usages:

build()

build(sample)

build(sample, method)

build(param)

Parameters:
sample2-d sequence of float, of dimension 1

The sample from which the distribution parameters are estimated.

methodint

An integer corresponding to a specific estimator method:

  • 0 : Local likelihood maximum estimator

  • 1 : Modified moment estimator

  • 2 : Method of moment estimator

  • 3 : Least squares method.

The default value is 0. It is stored in ResourceMap, key LogNormalFactory-EstimationMethod.

paramCollection of PointWithDescription

A vector of parameters of the distribution.

Returns:
distDistribution

The built distribution.

Notes

See the buildAsLogNormal() method.

buildAsLogNormal(*args)

Build the distribution as a LogNormal type.

Available usages:

buildAsLogNormal()

buildAsLogNormal(sample)

buildAsLogNormal(sample, method)

buildAsLogNormal(param)

Parameters:
sample2-d sequence of float, of dimension 1

The sample from which the distribution parameters are estimated.

methodint

An integer ranges from 0 to 2 corresponding to a specific estimator method:

  • 0 : Local likelihood maximum estimator

  • 1 : Modified moment estimator

  • 2 : Method of moment estimator

  • 3 : Least squares method.

Default value is 0. It is stored in ResourceMap, key LogNormalFactory-EstimationMethod.

paramCollection of PointWithDescription

A vector of parameters of the distribution.

Returns:
distLogNormal

The estimated distribution as a LogNormal.

Notes

In the first usage, the default LogNormal distribution is built.

In the second usage, the parameters are evaluated according the following strategy:

  • It first uses the local likelihood maximum based estimator.

  • It uses the modified moments based estimator if the resolution of (7) is not possible.

  • It uses the moments based estimator, which are always defined, if the resolution of (5) is not possible.

In the third usage, the parameters of the LogNormal are estimated using the given method.

In the fourth usage, a LogNormal distribution corresponding to the given parameters is built.

buildEstimator(*args)

Build the distribution and the parameter distribution.

Parameters:
sample2-d sequence of float

Data.

parametersDistributionParameters

Optional, the parametrization.

Returns:
resDistDistributionFactoryResult

The results.

Notes

According to the way the native parameters of the distribution are estimated, the parameters distribution differs:

  • Moments method: the asymptotic parameters distribution is normal and estimated by Bootstrap on the initial data;

  • Maximum likelihood method with a regular model: the asymptotic parameters distribution is normal and its covariance matrix is the inverse Fisher information matrix;

  • Other methods: the asymptotic parameters distribution is estimated by Bootstrap on the initial data and kernel fitting (see KernelSmoothing).

If another set of parameters is specified, the native parameters distribution is first estimated and the new distribution is determined from it:

  • if the native parameters distribution is normal and the transformation regular at the estimated parameters values: the asymptotic parameters distribution is normal and its covariance matrix determined from the inverse Fisher information matrix of the native parameters and the transformation;

  • in the other cases, the asymptotic parameters distribution is estimated by Bootstrap on the initial data and kernel fitting.

buildMethodOfLeastSquares(sample)

Build the distribution based on the least-squares estimator.

Parameters:
sample2-d sequence of float, of dimension 1

The sample from which the distribution parameters are estimated.

gammafloat, optional

The \gamma parameter estimate

Returns:
distLogNormal

The built distribution.

buildMethodOfLocalLikelihoodMaximization(sample)

Build the distribution based on the local likelihood maximum estimator.

Parameters:
sample2-d sequence of float, of dimension 1

The sample from which the distribution parameters are estimated.

Returns:
distLogNormal

The built distribution.

buildMethodOfModifiedMoments(sample)

Build the distribution based on the modified moments estimator.

Parameters:
sample2-d sequence of float, of dimension 1

The sample from which the distribution parameters are estimated.

Returns:
distLogNormal

The built distribution.

buildMethodOfMoments(sample)

Build the distribution based on the method of moments estimator.

Parameters:
sample2-d sequence of float, of dimension 1

The sample from which the distribution parameters are estimated.

Returns:
distLogNormal

The built distribution.

getBootstrapSize()

Accessor to the bootstrap size.

Returns:
sizeint

Size of the bootstrap.

getClassName()

Accessor to the object’s name.

Returns:
class_namestr

The object class name (object.__class__.__name__).

getName()

Accessor to the object’s name.

Returns:
namestr

The name of the object.

hasName()

Test if the object is named.

Returns:
hasNamebool

True if the name is not empty.

setBootstrapSize(bootstrapSize)

Accessor to the bootstrap size.

Parameters:
sizeint

The size of the bootstrap.

setName(name)

Accessor to the object’s name.

Parameters:
namestr

The name of the object.

Examples using the class

Fitting a distribution with customized maximum likelihood

Fitting a distribution with customized maximum likelihood