experiment_functions: Estimate the optimal posterior distribution of the data...
In xiangli2pro/hbcm: HBCM model for cluster analysis

heterogbcm_converge_qc_fix_hlambda

R Documentation

Estimate the optimal posterior distribution of the data column labels.

Description

'heterogbcm_converge_qc_fix_hlambda( )' uses given initial value of hlambda and hsigma, and keep the hlambda constant over iterations

'heterogbcm_logL( )' experiment function to compare the initial logL and last logL.

'heterogbcm_iterStep( )' experiment function with different update algorithm first update q2 and parameters Theta, then update q1 and parameters Theta.

'heterogbcm_noInitLabel( )' experiment function with no initial guess of the labels.

'heterogbcm_qcDiscrete( )' experiment function to update qc with discrete values.

Usage

heterogbcm_converge_qc_fix_hlambda(
  x,
  centers,
  tol,
  iter,
  iter_init = 3,
  labels,
  verbose = FALSE,
  hlambda
)

heterogbcm_logL(x, centers, tol, iter, iter_init = 3, labels, verbose = FALSE)

heterogbcm_iterStep(
  x,
  centers,
  tol,
  iter,
  iter_init = 3,
  labels,
  verbose = FALSE
)

heterogbcm_noInitLabel(
  x,
  centers,
  tol,
  iter,
  iter_init = 3,
  labels = NA,
  verbose = FALSE
)

heterogbcm_qcDiscrete(
  x,
  centers,
  tol,
  iter,
  iter_init = 3,
  labels,
  verbose = FALSE
)

Arguments

`x`	a numeric matrix data.
`centers`	an integer specifying the number of clusters.
`tol`	numerical tolerance of the iteration updates.
`iter`	number of iterations.
`iter_init`	iteration times of initial parameter estimation
`labels`	a vector specifying the cluster labels of the columns of x.
`verbose`	if TRUE, print iteration information.
`init_hlambda`	initial values for parameter vector Lambda.
`init_hsigma`	initial values for parameter vector Sigma.

Value

A list of values.

`omega`	estimated optimal group-correlation matrix.
`hlambda`	estimated optimal heterogeneous parameter Lambda.
`hsigma`	estimated optimal heterogeneous parameter Sigma.
`obj_logL_val`	vector of -logL from each iteration.
`qc`	estimated optimal posterior distribution of the column labels.
`cluster`	a vector of integers (from 1:k) indicating the cluster to which each column is allocated.