Estimation Models#

The package ships with parametric estimation models covering the most common univariate distributions. Each model can be selected by string name in MMDEstimator (e.g. model="gaussian-loc") or by passing a class instance directly. par_v denotes the variable parameter(s) that are optimised; par_c denotes constant parameter(s) that are held fixed. Click a model name for its full reference page.

regmmd.models.Gaussian

Gaussian \(\mathcal{N}(\mu, \sigma^2)\) with both mean and standard deviation estimated jointly.

regmmd.models.GaussianLoc

Gaussian \(\mathcal{N}(\mu, \sigma^2)\) with mean \(\mu\) estimated and standard deviation fixed.

regmmd.models.GaussianScale

Gaussian \(\mathcal{N}(\mu, \sigma^2)\) with standard deviation estimated and mean fixed.

regmmd.models.Beta

Beta \(\mathrm{Beta}(\alpha, \beta)\) with both shape parameters estimated jointly.

regmmd.models.BetaA

Beta \(\mathrm{Beta}(\alpha, \beta)\) with shape \(\alpha\) estimated and \(\beta\) fixed.

regmmd.models.BetaB

Beta \(\mathrm{Beta}(\alpha, \beta)\) with shape \(\beta\) estimated and \(\alpha\) fixed.

regmmd.models.Binomial

Binomial \(B(n, p)\) with success probability \(p\) estimated and number of trials \(n\) fixed.

regmmd.models.Gamma

Gamma distribution with both shape and rate estimated jointly.

regmmd.models.GammaShape

Gamma distribution with shape estimated and rate fixed.

regmmd.models.GammaRate

Gamma distribution with rate estimated and shape fixed.

regmmd.models.Poisson

Poisson distribution with rate \(\lambda\) estimated.

regmmd.models.Cauchy

Cauchy(loc, 1) — location-only parameterisation matching the R package.

regmmd.models.Dirac

1D Dirac mass at par_v.

regmmd.models.ContinuousUniformLoc

Continuous uniform on [c - L/2, c + L/2] — location estimated, length L fixed (par_c).

regmmd.models.ContinuousUniformUpper

Continuous uniform on [lower, upper] — upper bound estimated, lower bound fixed (par_c).

regmmd.models.ContinuousUniformLowerUpper

Continuous uniform on [lower, upper] — both bounds estimated.

regmmd.models.DiscreteUniform

Discrete uniform on \(\{1, 2, ..., N\}\) with N estimated by exact MMD minimisation.

regmmd.models.Geometric

regmmd.models.Pareto