get.center: Group clusters together until the main cluster contain the...

In KrasnitzLab/CNprep: Pre-process DNA Copy Number (CN) Data for Detection of CN Events

Group clusters together until the main cluster contain the minimum required ratio of data.

Description

The function groups clusters with the mean value closer to z zero together until the main cluster contain the minimum required ratio of data, as specified by the user.

Usage

get.center(emfit, minCenter)

Arguments

emfit

a list containing information about the current clusters obtained from a mixture model estimation: mu a numeric vector representing the mean for each component. If there is more than one component, the kth element is the mean of the kth component of the mixture model. pro a vector whose kth component is the mixing proportion for the kth component of the mixture model. If missing, equal proportions are assumed. z a numeric matrix whose [i,k]th entry is the probability that observation i in the test data belongs to the kth class. groups a matrix with the number of rows corresponding to the current number of clusters while the number of columns is corresponding to the initial number of clusters. The presence of 1 in position [i,k] indicates that the initial ith cluster is now part of the new kth cluster. ngroups a numeric, used as an integer, giving the final number of clusters. sigmasq a numeric vector giving the common variance for each component in the mixture model "E".

minCenter

a single numeric value between 0 and 1 specifying the minimal share of the central cluster in each profile.

Value

a list containing information about the current clusters obtained from a mixture model estimation:

• mu a numeric vector representing the mean for each component. If there is more than one component, the kth element is the mean of the kth component of the mixture model.

• pro a vector whose kth component is the mixing proportion for the kth component of the mixture model. If missing, equal proportions are assumed.

• z a numeric matrix whose [i,k]th entry is the probability that observation i in the test data belongs to the kth class.

• groups a matrix with the number of rows corresponding to the current number of clusters while the number of columns is corresponding to the initial number of clusters. The presence of 1 in position [k,i] indicates that the initial ith cluster is now part of the new kth cluster.

• ngroups a numeric, used as an integer, giving the final number of clusters.

• sigmasq a numeric vector giving the common variance for each component in the mixture model "E".

• center a numeric, used as an integer, indicating the cluster that has the mean closest to zero.

Author(s)

Alexander Krasnitz, Guoli Sun

Examples

## Create a list with mixture model estimation data containing 5 clusters
demoEM <- list()
demoEM[["mu"]] <- c(-0.23626, -0.08108, -0.02205, 0.03059, 0.24482)
demoEM[["pro"]] <- rep(0.2, 5)
demoEM[["z"]] <- matrix(data=c(1.19e-118, 2.81e-25, 5.87e-08, 9.99e-1,  
    1.86e-52, 2.03e-117, 9.19e-25, 1.02e-07, 9.99e-01, 1.92e-53, 1.00e+0, 
    1.34e-23, 1.72e-50, 1.08e-82, 6.45e-295, 1.00e+00, 1.39e-20, 2.51e-46, 
    1.67e-77, 1.47e-285, 8.86e-63, 1.21e-04, 9.99e-01, 1.89e-05, 7.93e-106,
    7.59e-60, 7.76e-04, 9.99e-01, 3.60e-06, 1.75e-109, 0.00e+0, 1.61e-147, 
    1.08e-98, 2.31e-63, 1.00e+0, 0.00e+0, 1.18e-147, 8.37e-99, 1.88e-63, 
    1.00e+0, 3.51e-75, 9.79e-01, 4.55e-08, 2.06-02, 2.14e-90, 7.07e-79,
    8.58e-01, 3.96e-09,  1.41e-01, 6.42e-86), ncol=5, byrow=TRUE)
demoEM[["groups"]] <- diag(x=1, nrow=5, ncol=5, names=TRUE)
demoEM[["ngroups"]] <- 5
demoEM[["sigmasq"]] <- rep(1.533e-3, 5)

## Group clusters until the minimum proportion of 40% of the data is in
## the main cluster. The main cluster being defined as the one closer to
## a value of zero.
result <- CNprep:::get.center(emfit=demoEM, minCenter=0.4)

## The result contain only 4 clusters as the clusters 3 and 4 have been
## grouped together to form a cluster that includes 40% of the data (group 3)
result$ngroups
result$pro

KrasnitzLab/CNprep documentation built on May 28, 2022, 8:32 p.m.