FitGoM: Run Grade of Membership (GoM) model with multiple starting...

In kkdey/CountClust: Clustering and Visualizing RNA-Seq Expression Data using Grade of Membership Models

Description

Fits grade of membership model FitGoM() to count data with multiple starting points and choose the best fit using BIC (Bayesian Information Criterion). the multiple starting points ensure that the output is more reliable.

Usage

FitGoM(data, K, tol = 100, num_trials = 1, options, path_rda = NULL,
  control = list())

Arguments

data	counts data N x G, with N, the number of samples along the rows and G, number of genes along columns.
K	the number of clusters to fit. Must be an integer.
tol	Tolerance value for GoM model absolute log posterior increase at successive iterations (set to 0.1 as default).
num_trials	The number of trials with different starting points used.
options	the measure used to choose best fit, either "BF" or "BIC" measures can be used. BF is more trustworthy, but BIC can be used for better model comparison.
path_rda	The directory path for saving the GoM model output. If NULL, it will return the output to console.
control	Control parameters. Same as topics() function of maptpx package. For example, verb controls information that are shown on the screen. verb=1 prints log2 Bayes factor. verb=2 prints log posterior at each iteration in addition to log2BF.

Value

Outputs the best GoM model fit output for cluster K and saves it at the directory path in path_rda if the latter is provided.

References

Matt Taddy. On Estimation and Selection for Topic Models. AISTATS 2012, JMLR W\&CP 22.

Pritchard, Jonathan K., Matthew Stephens, and Peter Donnelly. Inference of population structure using multilocus genotype data. Genetics 155.2 (2000): 945-959.

Examples

data("ex.counts")
out <- FitGoM(ex.counts, K=2, tol=100, num_trials=5,
                   control=list(tmax=100))

kkdey/CountClust documentation built on Jan. 17, 2021, 5:32 p.m.