R Graphical Manual

Browse All

Last data update: 2014.03.03

R: Constructing a named matrix of all two-way interactions.

maketwoway

R Documentation

Constructing a named matrix of all two-way interactions.

Description

maketwoway takes a set of pre-treatment covariates and returns all sensible two-way interactions.

Usage

maketwoway(X, wts = 1, center = TRUE, deletion = TRUE, threshold = 0.99999,
frame.meanPre, predict = FALSE)

Arguments

`X`	A matrix of pre-treatment covariates.
`wts`	An optional set of extrapolation weights.
`center`	Whether the columns should be centered before interacting.
`deletion`	Whether to automatically drop pairwise linearly dependent columns and columns of all 0's or 1's.
`threshold`	Threshold to drop correlated columns.
`frame.meanPre`	An argument to use when `maketwoway` is called within `predict.FindIt`.
`predict`	An argument to use when `maketwoway` is called within `predict.FindIt`.

Details

Useful for constructing a matrix for FindIt. Takes a matrix and returns a matrix of two-way interactions. A named matrix with an intercept, all two-way interactions and square terms are returned. If center=TRUE, interactions are formed by (1) centering, (2) scaling to sd 1, and (3) interacting. Non-sensible interactions, such as interacting dummy variables for male and female, as well as squared terms for dummy variables, are not returned. A rescaling vector is returned, so that the columns of the standardized matrix can be returned to their original scale.

Value

`X`	A matrix with all linear terms, squared terms, and two-way interactions from the original matrix.
`scale.X`	A vector of weights for reutrning X to its original scale.

Author(s)

Naoki Egami, Marc Ratkovic and Kosuke Imai.

Examples

#See the help page for FindIt() for an example.