Generates data from a low-rank DBN with factored A = U diag(alpha) U'
Usage
simulate_lowrank_dbn(
n = 30,
n_col = n,
p = 2,
time = 50,
r = 3,
sigma2 = 0.5,
tau_alpha2 = 0.1,
tauB2 = 0.05,
ar1_alpha = TRUE,
rho_alpha = 0.9,
seed = NULL,
return_truth = TRUE
)Arguments
- n
Number of sender actors
- n_col
Number of receiver actors (default: n)
- p
Number of relation types
- time
Number of time points
- r
Rank of the factorization
- sigma2
Process noise variance
- tau_alpha2
Variance for alpha factor innovations
- tauB2
Variance for B innovations
- ar1_alpha
Use AR(1) for alpha dynamics (default TRUE)
- rho_alpha
AR(1) persistence for alpha (default 0.9)
- seed
Random seed for reproducibility
- return_truth
If TRUE (default), include true parameters in output
Value
A list containing:
- Y
Observed ordinal data array
[n, n, p, time]- Z
Continuous latent values (use with
family = "gaussian")- Theta
True latent network state at each time point
- U
True orthogonal factor matrix
[n, r]- alpha
True factor trajectories
[r, time]- A
True time-varying sender influence
[n, n, time](reconstructed from U and alpha)- B
True time-varying receiver influence
[n_col, n_col, time]- M
True baseline mean array
[n, n_col, p]- sigma2, tau_alpha2, tauB2, r
True parameter values used in simulation
See also
dbn for model fitting, simulate_test_data for quick test data