runsugsvarsel: Internal function for SUGS with variable selection algorithm....

In ococrook/sugsvarsel: SUGS algorithm of Wang and Dunson (2011), SUGS VarSel algorithm Crook et al.

Description

Internal function for SUGS with variable selection algorithm. This function manages the random orderings and loops.

Usage

runsugsvarsel(mydata, featiter, clustiter, intfeatures, Model, mu_0 = NULL,
  lambda_0 = 0.01, nu_0 = NULL, S_0 = NULL, betaHat = c(0.01, 0.1, 1, 5,
  10, 15, 30, 50, 100), a = 10, b = 1, w = c(0.5, 0.5),
  BPPARAM = bpparam())

Arguments

mydata	A data matrix with rows as observations.
featiter	The number of iterations of variable selection
clustiter	The number of random ordering, for which to apply SUGS.
intfeatures	A binary matrix of the intial variable set, probably chosen using function pSelect. See documentation for pSelect for more details.
Model	The method used for Model select, either PML, ML or Both. If you select both the PML will be used to perform model selection.
mu_0	The mean hyperparameter, default is the column means of the data matrix.
lambda_0	The variance of the Guassian mean prior, the dafault value is 0.01.
nu_0	The degrees of freedom hyperparameter, the default value is 2 * (D + 2), where D is the number of variables.
S_0	The scale hyperparamter, the deault value is a tenth of the column variance of the data matrix.
betaHat	A grid of hyperparameters for the dirichlet concentration parameter, the default is c(1, 5, 15, 30, 50, 100).
a	The scale of the gamma prior for the dirichlet concentration parameter, the dafault value is 10.
b	The rate of the gamma prior for the dirichlet concentration parameter, the default value is 1.
w	The prior probability of a variable belong to the irrelevant or relevant partition. The vector must contain two entries the first entry being the probabiliy of being irreleavnt and the second being the probability of being relevant The default value is c(0.5,0.5).
BPPARAM	Support for parallel processing using the BiocParallel infrastructure. When missing (default), the default registered BiocParallelParam parameters are used. Alternatively, one can pass a valid BiocParallelParam parameter instance: SnowParam, MulticoreParam, DoparParam, ... see the BiocParallel package for details. To revert to the origianl serial implementation, use NULL.

Value

A matrix of cluster allocations for the number of iterations, a vector of either log PML, log ML or both, the number of clusters at each interations, the random ordering used and the last function output for these.

ococrook/sugsvarsel documentation built on May 27, 2019, 12:12 p.m.