iForm: Interaction Screening for Ultra-High Dimensional Data

In PennStateStatGen/iForm: Forward Selection Under Marginality Principle

Description

Extended variable selection approaches to jointly model main and interaction effects from high-dimensional data orignally proposed by Hao and Zhang (2014) and extended by Gosik and Wu (2016). Based on a greedy forward approach, their model can identify all possible interaction effects through two algorithms, iFORT and iFORM, which have been proved to possess sure screening property in an ultrahigh-dimensional setting.

Usage

iForm(formula, data, heredity = "strong", higher_order = FALSE)

Arguments

formula	an object of class "formula" (or one that can be coerced to that class): a symbolic description of the model to be fitted. The details of model specification are given under ‘Details’.
data	data.frame of your data with the response and all p predictors
higher_order	logical TRUE indicating to include order-3 interactions in the search (default FALSE)
strong	logical TRUE to use strong heredity or FALSE to use weak heredity (default TRUE)

Details

Runs the iFORM selection procedure on the dataset and returns a linear model of the final selected model.

Value

a summary of the linear model returned after the selection procedure

Author(s)

Kirk Gosik

See Also

lm

model.matrix

PennStateStatGen/iForm documentation built on May 8, 2019, 1:29 a.m.