Implements the sequential BART (Bayesian Additive Regression Trees) approach to impute the missing covariates. The algorithm applies a Bayesian nonparametric approach on factored sets of sequential conditionals of the joint distribution of the covariates and the missingness and applying the Bayesian additive regression trees to model each of these univariate conditionals. Each conditional distribution is then sampled using MCMC algorithm. The published journal can be found at < https://doi.org/10.1093/biostatistics/kxw009> Package provides a function, seqBART(), which computes and returns the imputed values.
This program is used for imputing missing covariates by the 'sequential BART' approach. Package provides a function, serBARTfunc, which computes and returns the imputed values.
The pacakge provides a function, seqBART(), to run the sequential BART model to find the missing covariates. The function takes as arguments 1. X, Covariates having the missing values.
Y, Response Variable.
Datatype, representing the type of covariate, continuous or binary,
Type, representing the type of missingness of the covaraites. It can take 3 values: 0 to represent covariates are MAR with MDM not depending on the response, and 1 or 2 to represent covariates are MAR with MDM depending on the response. If the response is continuous, use type=1 ( linear regression used for imputation), else if it is binary, use type=2 (logistic regression used for imputation).
Rest of the arguments are standard values for proper imputation. Defaults are provided.
sbart::seqBART(xx=Xcovariates, yy=Response, datatype=datatypeValues, type=1)
Changes and New Features in 0.1.1 (2018-05-01):
* comment out pragmas from Eigen code just like RcppEigen does, incorporate correct LICENSE and COPYRIGHT file for Eigen, and strip the libs on Linux to save space on CRAN Rodney Sparapani <[email protected]>