plotMSV-methods: Estimates the GPD and SPOmulti-source variability metrics and...

In carsaesi/EHRsourceVariability: Measurement and visualization of biomedical data heterogeneity across data sources

Description

Estimates the GPD and SPOmulti-source variability metrics and the corresponding source projection vertices from a matrix of probability distributions of different data sources

Usage

estimateMSVmetrics(msvMetrics, nBySource, idSource)

Arguments

msvMetrics	the output of estimateMSVmetrics
nBySource	number of individuals for each source
idSource	identifier for each source (character array)

Value

a plotly plot object

Examples

## Not run: 

library("PCAmixdata")

# We are going to estimate the MSV metrics and plot an MSV plot of the three first PCA coordinates of a dataset containing both numerical and categorical data
# We assume 'data' is a data.frame including numerical and categorical variables

# We get the indices of numerical and categorical data

quantidx = sapply(data,class) %in% c("numeric","integer")
qualiidx = sapply(data,class) %in% c("factor","character")

# We estimate a PCA projection using PCAmix for both numerical and categorical data
mca = PCAmix(X.quanti = NULL, X.quali = datasetVarsC2, ndim = 3, rename.level = TRUE, graph = FALSE)
coords = mca$ind$coord

# 'ID_SOURCE' contains the data source tag for each row in the data
# We get a kernel density estimation for the distributions of each source, removing those NULL estimations next

kdeData = by(coords[,1],ID_SOURCE, density, n = 100, from = min(coords[,1]), to = max(coords[,1]))
kdeData = lapply(kdeData,function(x) x$y)
kdeNull = sapply(kdeData,is.null)
kdeNotNull = kdeData[!kdeNull]
kdeDataNotNull = matrix(unlist(kdeNotNull), ncol = length(kdeNotNull), byrow = FALSE)
probMatrix = sweep(kdeDataNotNull, 2, colSums(kdeDataNotNull), FUN="/")

# We estimate the MSV metrics

msvMetrics = multisourcestability(probMatrix)
idSource = levels(ID_SOURCE)

nBySource = table(ID_SOURCE)

plotMSV(msvMetrics, nBySource, idSource)


## End(Not run)

carsaesi/EHRsourceVariability documentation built on June 6, 2019, 7:52 a.m.