Penalized Transformation Models in Liesel

Installation

The library can be installed from GitHub:

$ pip install git+https://github.com/liesel-devs/liesel-ptm.git#egg=liesel_ptm

Acknowledgements

Liesel-PTM is developed by Johannes Brachem with support from Paul Wiemann and Thomas Kneib at the University of Göttingen. As a specialized extension, Liesel-PTM belongs to the Liesel project. We are grateful to the German Research Foundation (DFG) for funding the development through grant 443179956.

API Reference

Model

PTMLocScale	Penalized transformation model for location and scale.
PTMLocScalePredictions	Posterior predictions for a penalized transformation model.
ShapePrior	Used to define the prior for the shape parameters of the transformation function.
Predictor	Organizes the terms of a structured additive predictor.

Covariate terms

LinearTerm	Linear function of one or more covariates.
StructuredAdditiveTerm	Term in a structured additive predictor.
PSpline	A P-spline with second-order random walk prior.
RandomIntercept	A random intercept with iid normal prior in noncentered parameterization.

Variable classes

VarWeibull	A variable with a Weibull prior.
VarInverseGamma	A variable with an inverse gamma prior.
VarHalfCauchy	A variance parameter with a half Cauchy prior on its square root.
ScaleWeibull	A variable with a Weibull prior on its square.
ScaleInverseGamma	A variable with an inverse gamma prior on its square.
ScaleHalfCauchy	A scale parameter with a half Cauchy prior.
SymmetricallyBoundedScalar	Class for defining a scalar \(\omega\) with symmetrically bounded support around 1.
TransformedVar	Class for defining a possibly transformed variable.

Data generation

PTMLocScaleDataGen	Draws random samples from a location-scale transformation model.
sample_shape	Draws a random sample of the shape parameters \(\boldsymbol{\delta}\) from a first order random walk prior.

Helpers

nullspace_remover	Constructs a reparameterization matrix fo removing the nullspace from a penalty matrix of a structured additive predictor term.
diffpen	A P-spline penalty matrix based on diff-order differences.
sumzero_term	Matrix Z for reparameterization for sum-to-zero-constraint of a structured additive term.
sumzero_coef	Matrix Z for reparameterization for sum-to-zero-constraint of coefficients.
bspline_basis	Vectorized B-spline basis function evaluation.
kn	Create equidistant knots for B-Spline of the specified order.

Indices and tables

• Index

• Module Index

• Search Page