estimate_densities: MLE for Normal Mixture Model in 'hetset' Object
In ZytoHMGU/hetset: Identification of Heterogeneous Subsets in Data

Description Usage Arguments Details Value Author(s) See Also Examples

Densities for the two components of a normal mixture model are calculated (MLE) for a hetset object. The groups labels and a set of selected features have to be specified.

1	estimate_densities(H)

`H`	hetset object with non-empty set of selected features and a partitioning given in the sample data.

For a set of selected features (in H@metadata$slf) and a partitioning (in H$prt), the maximum likelilhood esetimator (MLE) for the paramters of a two-component normal mixture model are calculated and written to the metadata of H.

For single features, mean and variance are written to H@metadata$prm.a$mean, H@metadata$prm.a$cov, ... , for higher dimensions, mean-vector and covariance matrix are stored.

A valid partitioning is a factor with levels "A" and "B" as returned from reassign_samples or scan_hetset.

Returns the updated SummarizedExperiment object with updated parameters for the mixture model

`metadata$prm.full`	list with mean vector and covariance matrix for the selected features
`metadata$prm.A`	list with mean vector and covariance matrix for subpopulation A
`metadata$prm.B`	list with mean vector and covariance matrix for subpopulation B
`metadata$prp.A`	mixture coefficient indicating the relative number of samples assigned to subpopulation A

Daniel Samaga

scan_hetset

A <- matrix(data = rnorm(n = 1000,mean = 1,sd = 1),ncol = 20)
B <- matrix(data = rnorm(n = 1000,mean = 5,sd = 1),ncol = 20)
Hds <- hetset(D = cbind(A,B))
rm(A,B)
Hds$prt <- as.factor(sample(c("A","B"),ncol(Hds),TRUE))
Hds@metadata$slf <- c("F5","F15")
Hds <- estimate_densities(Hds)
print(Hds@metadata)

data("TCGA_HNSCC_expr")
H@metadata$slf <- sample(H@NAMES,3,FALSE)
H$prt <- as.factor(sample(c("A","B"),ncol(H),TRUE))
H <- censor_data(H)
H <- estimate_densities(H)
print(H@metadata)