Regression Models#

The regression models follow the same conventions as the estimation models (see Estimation Models). Select a model by string name in MMDRegressor (e.g. model="linear-gaussian-loc") or pass a class instance directly. Click a model name for its full reference page.

regmmd.models.LinearGaussian

Linear regression \(Y \mid X \sim \mathcal{N}(X^\top\beta, \phi)\) with both regression coefficients \(\beta\) and noise variance \(\phi\) estimated jointly.

regmmd.models.LinearGaussianLoc

Linear regression \(Y \mid X \sim \mathcal{N}(X^\top\beta, \phi)\) with regression coefficients \(\beta\) estimated and noise variance \(\phi\) fixed.

regmmd.models.Logistic

Logistic regression \(Y \mid X \sim \mathrm{Bernoulli}(\sigma(X^\top\beta))\) with regression coefficients estimated.

regmmd.models.GammaRegression

Gamma regression with mean \(\mu = \exp(X^\top\beta)\); coefficients \(\beta\) and shape parameter estimated jointly.

regmmd.models.GammaRegressionLoc

Gamma regression with mean \(\mu = \exp(X^\top\beta)\); coefficients \(\beta\) estimated and shape parameter fixed.

regmmd.models.PoissonRegression

Poisson regression \(Y \mid X \sim \mathrm{Poisson}(\exp(X^\top\beta))\) with regression coefficients estimated.

regmmd.models.BetaRegression

Beta regression where both beta and the log-precision are estimated.

regmmd.models.BetaRegressionLoc

Beta regression with the precision phi fixed and only beta estimated.