abess (version 0.4.11)

generate.bmn.data: Generate a dataset from binary markov network

Description

Generate a dataset with binary variables upon the interaction matrix. If the interaction matrix is not provided, and then, an interaction matrix would be generated.

Usage

generate.bmn.data(
  n,
  p,
  type = 1,
  seed = NULL,
  graph.seed = NULL,
  theta = NULL,
  beta = 0.7,
  degree = 3,
  alpha = 0.4,
  method = "freq"
)

Value

a list that includes three components: data, weight, theta. data is an n-by-p matrix that record samples; weight: an n-dimensional vector that records the frequency of each row of data; theta is the interaction matrix used for generating samples.

Arguments

n

The number of observations.

p

The number of predictors of interest.

type

a numeric value that indicates the type of an interaction matrix

seed

Seed to be used to reproduce dataset

graph.seed

Seed used to reproduce the interaction matrix

theta

An user specified interaction matrix. It should be a symmetric matrix.

beta

a numeric value that specifies the remaining signal of the interaction matrix

degree

an integer value that specifies the maximum node degree in the interaction matrix

alpha

a numeric value that specifies the minimum signal of the interaction matrix

method

the method used for generating samples. One is the "freq", another is it "gibbs". The former is more suitable when p<18, while the later is more suitable when p > 18.

Examples

library(abess)
p <- 16
n <- 1e3
train <- generate.bmn.data(n, p, type = 7, graph.seed = 1, seed = 1, beta = 1.3, alpha = 0.2)
res <- slide(train[["data"]], train[["weight"]], tune.type = "gic")