In LilithF/doMIsaul: Do Multiple Imputation-Based Semi-Supervised and Unsupervised Learning

knitr::opts_chunk$set(
  collapse = TRUE,
  comment = "#>",
  fig.path = "man/figures/README-",
  out.width = "100%"
)

doMIsaul

Overview

The goal of \code{doMIsaul} package is to provide functions to perform unsupervised and semisupervised learning for an incomplete dataset.

Installation

You can install the development version from GitHub with:

# install.packages("devtools")
devtools::install_github("LilithF/doMIsaul")

Example

This is a basic example which shows you how to perform unsupervised learning for an incomplete dataset:

library(doMIsaul)
data(cancer, package = "survival")
cancer$status <- cancer$status - 1
cancer <- cancer[, -1]

set.seed(1243)
res.unsup <- 
  unsupMI(data = list(cancer), Impute = "MImpute_surv", Impute.m = 10,
          cleanup.partition = TRUE, return.detail = TRUE)

cancer$part_unsup <- res.unsup$Consensus

plot_MIpca(res.unsup$Imputed.data, 1:228, color.var = cancer$part_unsup,
           pca.varsel = c("age", "sex", "ph.ecog", "ph.karno", "pat.karno",
                          "meal.cal",  "wt.loss"))

plot_boxplot(data = cancer, partition.name = "part_unsup",
             vars.cont = c("age", "meal.cal", "wt.loss"),
             unclass.name = "Unclassified", include.unclass = FALSE)

plot_frequency(data = cancer, partition.name = "part_unsup",
               vars.cat = c("sex", "ph.ecog"))

This is a basic example which shows you how to perform semisupervised learning for an incomplete dataset with a survival outcome:

## With imputation included
set.seed(345)
res.semisup <- 
  seMIsupcox(X = list(cancer[, setdiff(colnames(cancer), "part_unsup")]),
             Y = cancer[, c("time", "status")],
             Impute = TRUE, Impute.m = 10, center.init = TRUE,
             nfolds = 10, center.init.N = 50, 
             cleanup.partition = TRUE, return.detail = TRUE)
# This is an example, a larger value for center.init.N is recommended.

cancer$part_semisup <- res.semisup$Consensus[[1]]

plot_MIpca(res.semisup$Imputed.data, NULL, color.var = cancer$part_semisup,
           pca.varsel = c("age", "sex", "ph.ecog", "ph.karno", "pat.karno",
                          "meal.cal",  "wt.loss"))

plot_boxplot(data = cancer, partition.name = "part_semisup",
             vars.cont = c("age", "meal.cal", "wt.loss"),
             unclass.name = "Unclassified", include.unclass = TRUE)

plot_frequency(data = cancer, partition.name = "part_semisup",
               vars.cat = c("sex", "ph.ecog"))

Reference publications

You may find more details on the methods implemented in this package in the associated publications:

• Unsupervised MI learning: Faucheux L, Resche-Rigon M, Curis E, Soumelis V, Chevret S., Clustering with missing and left-censored data: A simulation study comparing multiple-imputation-based procedures. Biometrical Journal. 2021; 63: 372– 393. https://doi.org/10.1002/bimj.201900366
• Semisupervised MI learning for a survival outcome: Faucheux L, Soumelis V, Chevret S., Multiobjective semisupervised learning with a right-censored endpoint adapted to the multiple imputation framework. Biometrical Journal. 2021; 1– 21. https://doi.org/10.1002/bimj.202000365

LilithF/doMIsaul documentation built on Dec. 17, 2021, 12:08 a.m.