################################################################################
################################################################################
Current folder contains the sources of R-package 'imputeDepth' and files required to reproduce all the experiments and data-connected figures of the manuscript 'Nonparametric imputation by data depth' (2017) by Pavlo Mozharovskyi, Julie Josse, and Francois Husson. Detail descriptions for utilisation of the files follow.
Folder 'imputeDepth' as well as the archive 'imputeDepth_0.1.0.tar.gz' contain sources of the R-package 'imputeDepth', whose functions implement the imputation methods from the manuscript, and which are used in the experiments.
Folder 'data' contains files of the data sets 'banknotes.dat', 'bloodtransfusion_gp.dat', 'cows.dat, 'glass.dat'; see the manuscript and references therein for details.
File 'impute.functions.R' contains routines for competing imputation methods implemented by the authors and a function for introducing of missing data into a complete data set.
Files 'exp.impute.*.R' contain R-sources of the experiments, namely:
-
files 'exp.impute.StudentT_MCAR.R', 'exp.impute.StudentToutl_MCAR.R', 'exp.impute.StudentT_MAR.R', 'exp.impute.lowrank.R', 'exp.impute.large.R', 'exp.impute.skewed.R', 'exp.impute.moon.R' contain scripts for simulation experiments with missing completely at random (MCAR), MCAR with outliers, missing at random (MAR), lowrank models, contamination in higher dimension, skewed normal distribution and distribution wiht non-convex support, respectively, see Section 4.2.
-
files 'exp.impute.banknotes.R', 'exp.impute.bloodtransfusion.R', 'exp.impute.cows.R', 'exp.impute.glass.R' contain scripts for real-data experiments; see Section 5.3. The data sets are supposed to be in the folder 'data', if not the path in the R-file should be changed accordingly.
-
file 'exp.impute.quantile.R' contains the R-script for reproducing the quantile experiment for multiple imputation; see Section 5.3.
-
file 'exp.impute.multiple.R' contains the R-script for reproducing the complete multiple imputation experiment for a regression model, see Section 5.2.
Files 'plot.intro.R', 'plot.impute.skewed.R', 'plot.impute.moon.R', 'plot3d.optim.R', 'plot.one.density.R' , 'plot.impute.Tukey.R' reproduce Figures 1 and 2 (Section 1), Figures 4 and 5 (Section 4.2.6), Figures 1--3 (supplement), respectively.
File 'plot3d.realdata.R' produces pairs() and pairs3d() plots used to get intuition by visualising the real data.