10. Copula factories¶

class lsqfitgp.copula.CopulaFactory(name, *params)¶

Abstract base class for copula factories.

Class to represent transformations in gvar BufferDicts, i.e., Gaussian copulas.

Parameters:

namestr: The unique name assigned to the transformation using gvar.BufferDict.add_distribution().
*paramsscalars: The parameters of the distribution.

Examples

>>> copula = gvar.BufferDict({
...     'transf(x)': lgp.copula.beta('transf', 1, 1),
...     'transf_2(y)': lgp.copula.beta('transf_2', 3, 5),
... })
>>> copula['x']
0.50(40)
>>> copula['y']
0.36(18)

abstract static invfcn(x, *params)¶: \(F^{-1}(\Phi(x))\), i.e., maps a Normal variable to a variable with the desired marginal distribution. Jax-traceable.

class lsqfitgp.copula.beta(name, alpha, beta)¶: https://en.wikipedia.org/wiki/Beta_distribution

class lsqfitgp.copula.invgamma(name, alpha, beta)¶: https://en.wikipedia.org/wiki/Inverse-gamma_distribution