nm_fit: Log-ratio normal-multinomial parameters estimation.

In mcomas/normalmultinomial: Normal-Multinomial Distribution

Description

The parameters mu and sigma are expressed with respect basis given by function @ilrBase.

Usage

nm_fit(X, eps = 0.001, nsim = 1000, parallel.cluster = NULL,
  max.em.iter = 100, min.em.iter = 10, expected = TRUE, verbose = FALSE,
  init.method = "dm", development = FALSE)

Arguments

X	count data set
eps	maximum error accepted on the last EM-step
nsim	number of simulations used in the E-step
parallel.cluster	parallel Socket Cluster created with function @makeCluster
max.em.iter	maximum number of steps allowed in the EM-algorithm
min.em.iter	minimum number of steps before allowing to double the number of iterations
expected	if TRUE the expected probabilities are returned (default:TRUE)
verbose	show information during estimation
init.method	how to initiate the method (default 'dm' initiating with dirichlet-multinomial estimation). Another options are: 'aitchison', 'bootstrap'. Initiating with aproximation from raw data and using bootstrapping respectively. For stability, we recommend to use 'dm' method.
development	show diagnostic messages

Value

A list with parameters mu and sigma and the number of iterations before convergence

Examples

X = rnormalmultinomial(100, 100, rep(0,4), diag(4))
nm_fit(X, verbose = T)

mcomas/normalmultinomial documentation built on May 22, 2017, 1:14 a.m.