Generate 10-variable sample data for Direct LiNGAM

Generates a sample dataset with a known causal structure. The true causal structure is: x3 -> x0 (coef = 3.0) x3 -> x2 (coef = 6.0) x3 -> x9 (coef = 7.0) x0 -> x1 (coef = 3.0) x0 -> x5 (coef = 4.0) x0 -> x4 (coef = 8.0) x0 -> x7 (coef = 3.0) x2 -> x1 (coef = 2.0) x2 -> x4 (coef = -1.0) x2 -> x8 (coef = 0.5) x1 -> x6 (coef = 2.0) x5 -> x8 (coef = 2.0) x4 -> x7 (coef = 1.5) x6 -> x9 (coef = 1.0)

Usage

generate_lingam_sample_10(n = 1000L, seed = 42L, noise_dist = "uniform")

Arguments

n

number of samples (default: 1000)

seed

random seed (default: 42)

noise_dist

error term distribution "uniform" : Uniform(0, 1) - default, non-Gaussian (LiNGAM works well) "gaussian" : Normal(0, 1) - LiNGAM may fail "lognormal" : Log-normal(0, 1) - skewed, non-Gaussian "exponential" : Exponential(1) - skewed, non-Gaussian "t3" : t-distribution (df=3) - heavy tails

Value

list(data, true_adjacency)

Examples

# Non-Gaussian (LiNGAM works well)
X_nongauss <- generate_lingam_sample_10(noise_dist = "uniform")
result <- lingam_direct(X_nongauss$data, reg_method = "ols")
result$causal_order
#>  [1]  4  3  1  5  6  8  2  9  7 10

# Gaussian (LiNGAM may fail)
X_gauss <- generate_lingam_sample_10(noise_dist = "gaussian")
result <- lingam_direct(X_gauss$data, reg_method = "ols")
result$causal_order
#>  [1] 10  2  7  3  9  4  6  5  8  1