Learn R Programming

BMSC (version 0.2.1)

Bayesian Model Selection under Constraints

Description

A Bayesian regression package supporting constrained coefficient estimation and variable selection using Stan. This includes a robust variable selection algorithm by a horseshoe prior () that finds the optimal model considering main effects, interactions as well as powers of given variables under potential parameter constraints.

Copy Link

Version

Install

install.packages('BMSC')

Monthly Downloads

9

Version

0.2.1

License

GPL-3

Maintainer

Marcus Gross

Last Published

August 2nd, 2019

Functions in BMSC (0.2.1)

Extract model names from model objects

Turn character vector into formula, return error if not possible

addInteractionToVars

Add interactions of a specific order to a vector of variables

ConstrainedLinReg-class

S4 class for constrained linear regression models

Sort a vector and collapse elements together using ":"

show,ConstrainedLinReg-method

Print constraint estimation model

makeInteractions

Add all interactions up to a desired order

Prepare data to plot model fit

Plot errors of all models

Add exponent to a vector of variables

Model selection algorithm for constrained estimation

Create polynomial of degree maxExponent from variable names

Plot model errors with errorbars

print.ConstrainedLinReg

Print constraint estimation model

createFormulaInternal

Create formula with interactions and polynomials if all checks in createFormula have passed

handleMissingData

Exclude rows with missing values

Extract variable name from polynomial expression

Extract beta matrix from ConstrainedLinReg model

Get Best Model after Models Selection

Create a formula with interactions and polynomials up to a desired order

prepDatForPredict

Exclude rows with missing data on predictor variables

predict,ConstrainedLinReg-method

Compute predictions from constraint estimation model

Prepare colour vector