InteractionProb: Calculate posterior inclusion probabilities for an...

In jantonelli111/NLinteraction: Bayesian variable selection in multivariate semi-parametric regression models

Description

This function takes a fitted NLint object and can return the posterior probability that any given set of variables are jointly interacting with each other.

Usage

InteractionProb(NLmod, Xsub)

Arguments

NLmod	The fitted NLint object
Xsub	The exposures for which you want to evaluate whether there is an interaction

Value

A number indicating the posterior probability that the exposures in Xsub are jointly interacting together. If Xsub is a scalar then the marginal posterior inclusion probability is returned, while if Xsub is a vector of length 2, the two way interaction is returned and so on.

Examples

n = 200
p = 10
pc = 1

sigma = matrix(0.3, p, p)
diag(sigma) = 1
X = rmvnorm(n, mean=rep(0,p), sigma = sigma)

C = matrix(rnorm(n*pc), nrow=n)

TrueH = function(X) {
  return(0.5*(X[,2]*X[,3]) - 0.6*(X[,4]^2 * X[,5]))
}

Y = 5 + C + TrueH(X) + rnorm(n)

NLmod = NLint(Y=Y, X=X, C=C)

## Find 2 way interaction probability between exposure 4 and 5
InteractionProb(NLmod=NLmod, Xsub=c(4,5))

jantonelli111/NLinteraction documentation built on Aug. 7, 2020, 10:02 p.m.