Simultaneously Impute the Missing and Censored Values.

mvnimpute

The goal of mvnimpute package is to implement multiple imputation for the multivariate data with both missing and censored values (a single variable can have both missing and censored values simultaneously; or it can have either only missing or censored values). An example of application of this package is for imputing the NHANES laboratory measurement data that are subject to both missing values and limits of detection (LODs).

Installation

For Windows users, the Rtools for building R packages has to be installed according to your R version from https://cran.r-project.org/bin/windows/Rtools/history.html.

From GitHub

NOTE: Some of the packages that this package depends on may require the latest version of R, it is recommended to update your R software to the latest version. The development version of the mvnimpute package can be installed from GitHub with:

For first-time users

# install the development package devtools for installing packages from GitHub
install.packages("devtools")

# install mvnimpute package from GitHub
devtools::install_github("hli226/mvnimpute")

You have to install the development package devtools for installing packages from GitHub. The packages that mvnimpute depends on will be automatically downloaded and installed.

Basic functions

It has 9 functions including

data.generation: generates multivariate normal data with missing and censored values.

visual.plot: draws plot showing percentages of missing, censored, and observed values.

marg.plot: draws marginal density plot for each variable.

multiple.imputation: multiply imputes data with missing and censored values.

conv.plot: draws convergence plot of the parameters from the multiple imputation.

avg.plot: draws convergence plot of the averaged values of the parameters from the multiple imputation.

acf.calc: calculates the autocorrelation values and draws ACF plots.

nhanes.dat: A subset of the 1999-2004 NHANES data with selected variables including diastolic blood pressure, gender, age and body mass index.

simulated.dat: A simulated dataset of sample size 200 with missing and left censored values.

Acknowlegements

This package is based on the work supported by the National Institute of Environmental Health Sciences (NIEHS) under grant 1R01ES028790.