BayesPen.refit: Bayesian Penalized Credible Regions Solution Path Refit
In AnderWilson/BayesPen: Bayesian Penalized Credible Regions

Description Usage Arguments Details Value Author(s) References See Also Examples

Refits the solution path given by BayesPen.

1	BayesPen.refit(y, x, fit, joint, max.refit, ...)

`y`	A n-vector of responses. If fit is a list from BayesPen.lm or BayesPen.lm.confounders then y is not required.
`x`	A n x p design matrix that includes all potential covariates. In the confounder selection case this includes the exposures and confounders, i.e. cbind(x,u). If fit is a list from BayesPen.lm or BayesPen.lm.confounders then x is not required.
`fit`	A list returned from BayesPen, BayesPen.lm, or BayesPen.lm.confounders.
`joint`	For variable selection this indicates if the joint or marginal solution path should be used. Joint must be TRUE for confounder selection.
`max.refit`	The maximum number of models to be refit.
`...`	These are additional terms passed to glm to refit a glm other than linear. The default is linear.

This refits each model in the solution path with the frequentist model using the glm function.

`coefs`	A matrix of regression coefficients for each model in the solution path. The regression coefficients for parameters omitted from a model are set to 0.
`SSE`	SSE of each refitted model.
`dev`	Deviance of each refitted model.
`df`	Error degrees of freedom from each refitted model.
`joint`	Returns the logical joint.

Ander Wilson, Howard D. Bondell, and Brian J. Reich

Bondell, H. D. and Reich, B. J. (2012). Consistent high-dimensional Bayesian variable selection via penalized credible regions. J. Am. Statist. Assoc. 107, 1610-1624.

Wilson A., Reich B. J. (2014). Confounder selection via penalized credible regions. Biometrics 70: 852-861.

BayesPen

######################
#Variable Selection
set.seed(1234)
dat <- SimExample(500,model="BR1")
X <- dat$X
y <- dat$y

#fit the full model assuming flat priors on beta
betahat <- solve(t(X)%*%X) %*% t(X) %*% y
cov <- solve(t(X)%*%X) * sum((X%*%betahat-y)^2)/(length(y)-length(betahat))

#find solution path
fit.BayesPen <- BayesPen(beta=betahat, beta_cov=cov)

#refit the model
refit <- BayesPen.refit(y,X,fit.BayesPen)

#plot it
BayesPen.plot(refit)