Generate sample data from a VARMA-LiNGAM model

Generates a 3-variable time series following a VARMA(1,1)-LiNGAM model with a strictly acyclic instantaneous structure B0, a lag-1 autoregressive matrix Phi1, a lag-1 moving-average matrix Theta1, and non-Gaussian (uniform) errors. The reduced-form recursion is x(t) = Phi1 x(t-1) + n(t) + Theta1 n(t-1) with n(t) = (I - B0)^{-1} e(t).

Usage

generate_varmalingam_sample(n = 1000, seed = NULL)

Arguments

n

number of time points to return (after burn-in)

seed

random seed (NULL allowed)

Value

list with data (data frame, n x 3), the reduced-form true matrices true_B0, true_phi1, true_theta1, and the structural-form counterparts true_psi1 = (I - B0) Phi1 and true_omega1 = (I - B0) Theta1 (I - B0)^{-1} for comparison against lingam_varma() estimates

Examples

sample <- generate_varmalingam_sample(n = 500, seed = 1)
head(sample$data)
#>           x0         x1         x2
#> 1  0.1735288  1.2799008 -1.2155499
#> 2 -0.4072948  1.2166111 -1.4989248
#> 3 -0.9548545  0.5966513 -1.4501052
#> 4  0.2940208  1.0468406 -0.1676003
#> 5  0.5680147  1.1038743  0.7978025
#> 6 -0.1768722 -0.7770469  1.5696898