Title: | Clustering Data with Non-Ignorable Missingness using Semi-Parametric Mixture Models |
---|---|
Description: | Clustering of data under a non-ignorable missingness mechanism. Clustering is achieved by a semi-parametric mixture model and missingness is managed by using the pattern-mixture approach. More details of the approach are available in Du Roy de Chaumaray et al. (2020) <arXiv:2009.07662>. |
Authors: | Marie Du Roy de Chaumaray [aut], Matthieu Marbac [aut, cre, cph] |
Maintainer: | Matthieu Marbac <[email protected]> |
License: | GPL (>= 2) |
Version: | 1.1.0 |
Built: | 2024-12-08 06:58:25 UTC |
Source: | CRAN |
Clustering method to analyze continuous or mixed-type data with missingness. The missingness mechanism can be non ignorable. The approach considers a semi-parametric mixture model.
Package: | MNARclust |
Type: | Package |
Version: | 1.1.0 |
Date: | 2021-12-01 |
License: | GPL-3 |
LazyLoad: | yes |
All the patients suffered heart attacks at some point in the past. Some are still alive and some are not. The survival and still-alive variables, when taken together, indicate whether a patient survived for at least one year following the heart attack.
A data frame with 132 observations on 13 variables (more details on this data set are presented in http://archive.ics.uci.edu/ml/datasets/Echocardiogram).
This data set arise from the UCI machine learning repository (more details on this data set are presented http://archive.ics.uci.edu/ml/datasets/Echocardiogram)
Salzberg, S. (1988). Exemplar-based learning: Theory and implementation (Technical Report TR-10-88). Harvard University, Center for Research in Computing Technology, Aiken Computation Laboratory (33 Oxford Street; Cambridge, MA 02138).
data(echo)
data(echo)
Clustering method to analyze continuous or mixed-type data with missingness. The missingness mechanism can be non ignorable. The approach considers a semi-parametric mixture model.
MNARcluster( x, K, nbinit = 20, nbCPU = 1, tol = 0.01, band = band.default(x), seedvalue = 123 )
MNARcluster( x, K, nbinit = 20, nbCPU = 1, tol = 0.01, band = band.default(x), seedvalue = 123 )
x |
matrix used for clustering |
K |
number of components |
nbinit |
number of random starting points |
nbCPU |
number of CPU used for parallel computing (only Unix and Linux systems are allowed) |
tol |
stopping rule |
band |
bandwidth (numeric vector). |
seedvalue |
value of the seed (used to set the initializations of the MM algorithm) |
Returns a list containing the proportions (proportions), matrix of probabilities of missngness (rho), the posterior probabilities of classification (classproba), the partition (zhat) and the logarithme of the smoothed-likelihood (logSmoothlike)
Clustering Data with Non-Ignorable Missingness using Semi-Parametric Mixture Models, Marie Du Roy de Chaumaray and Matthieu Marbac <arXiv:2009.07662>.
set.seed(123) # Data generation ech <- rMNAR(n=100, K=2, d=4, delta=2, gamma=2) # Clustering res <- MNARcluster(ech$x, K=2) # Confusion matrix between the estimated and the true partiion table(res$zhat, ech$z)
set.seed(123) # Data generation ech <- rMNAR(n=100, K=2, d=4, delta=2, gamma=2) # Clustering res <- MNARcluster(ech$x, K=2) # Confusion matrix between the estimated and the true partiion table(res$zhat, ech$z)
Generation of data set to perform the simulation presented in Section 4.1 of Du Roy de Chaumaray (2020)
rMNAR( n, K, d = 3, delta = 3, gamma = 1, law = "gauss", linkmissing = "logit-X" )
rMNAR( n, K, d = 3, delta = 3, gamma = 1, law = "gauss", linkmissing = "logit-X" )
n |
sample size (numeric of length 1) |
K |
number of clusters (numeric of length 1) |
d |
number of variables (numeric of length 1) |
delta |
tuning parameter to define the rate of misclassification (numeric of length 1) |
gamma |
tuning parameter to define the rate of missingness (numeric of length 1) |
law |
specifies the distribution of the variables within components (character that must be equal to gauss, student, laplace or skewgauss) |
linkmissing |
specify the missingness mechanism (character that must be equal to MCAR, logit-Z, logit-X or censoring) |
rMNAR returns a list containing the observed data (x), the true cluster membership (z), the complete data (xfull), the cluster membership given by the Baye's rule (zhat), the empirical rates of misclassification (meanerrorclass) and missngness (meanmiss).
Clustering Data with Non-Ignorable Missingness using Semi-Parametric Mixture Models, Marie Du Roy de Chaumaray and Matthieu Marbac <arXiv:2009.07662>.
set.seed(123) # Data generation ech <- rMNAR(n=100, K=3, d=3, delta=2, gamma=1) # Head of the observed data head(ech$x) # Table of the cluster memberships table(ech$z) # Empirical rate of misclassification ech$meanerrorclass # Empirical rate of missingness ech$meanmiss
set.seed(123) # Data generation ech <- rMNAR(n=100, K=3, d=3, delta=2, gamma=1) # Head of the observed data head(ech$x) # Table of the cluster memberships table(ech$z) # Empirical rate of misclassification ech$meanerrorclass # Empirical rate of missingness ech$meanmiss