generate_data: Simulate data from the cFIT model

In pengminshi/cFIT: cFIT (common Factor Space Integration & Transfer)

Description

Simulate

X_j = H_jWΛ_j + 1_{n_j}b_j +E_j, j = 1,…, ntask

. The nonnegative matrix is generated as the cluster centers given number of clusters K. H_j is generated as the binary membership matrix, where the labels are generated from a Dirichlet distribution with parameter alpha. Distortion lambda and shift b are generated from truncated normal distribution, Noise matrix E is generate with each entry from iid normal distribution.

Usage

generate_data(
  n,
  ntask,
  K,
  p,
  alpha = NULL,
  sig = 1,
  cl.sep = 1,
  batch.effect.sig = 0.1
)

Arguments

n	number of data point per dataset
ntask	number of batches
K	number of clusters
p	number of genes
alpha	parameter for Dirichilet distribution used to generate the labels (default 10, representing equal cluster size. smaller alpha corresponds to more unbalanced types)
sig	with cluster variance
cl.sep	Cluster center separation, the higher the clusters are well separated
batch.effect.sig	batch effect variance, higher the large batch effects are

Value

A list of generated data,

X.list

a list of n-by-p expression matrix

H.list

a list of n-by-K binary membership matrix

lambda.list

a list of length-p vectors of per-dataset scaling

b.list

a list of length-p vectors of per-dataset shift

E.list

a list of noise matrix

label.list

a list of length-n vectors of cluster labels

W

p-by-K common factor matrix

pengminshi/cFIT documentation built on July 11, 2021, 11:12 p.m.