R/calSI.R

In GaoLabXDU/CEPICS: CEPICS: a comparison and evaluation platform for integration methods in cancer subtyping

#'Calculate the Silhouette Coefficient of the result of a clustering algorithm.
#'
#'Calculate the silhouette coefficient of each sample, and take the average value.
#'
#'
#' @param x A data matrix with columns as features (e.g. genes) and rows as samples (e.g, patients).
#' @param a A vector that contains the cluster label of each sample.
#' @param type Choose from 0, 1 or 2, which means the data matrix x is the raw data matrix, similarity matrix, and distance matrix respectively.
#' @return A real number that means the silhouette coefficient of the cluster.
#' @examples
#' #sample one
#' s1=c(3,1,6,4,8)
#' #sample two
#' s2=c(6,7,1,3,4)
#' #sample three
#' s3=c(2,8,7,2,3)
#' label=c(1,1,3)
#'
#' x=as.matrix(rbind(s1,s2,s3))
#'
#' #the first parameter is row data
#' calSI(x,label,0)
#'
#' #the first parameter is similarity matrix
#' calSI(1/dist(x),label,1)
#'
#' #the first parameter is distance matrix
#' calSI(dist(x),label,2)
#' @export
calSI <- function(x,a,type=0)
{
  if(type==0)
    return(mean(silhouette(a,dist(x))[,3]))
  else if(type==1)
    return(mean(silhouette(a,as.dist(1/x))[,3]))
  else if(type==2)
    return(mean(silhouette(a,as.dist(x))[,3]))
}