simulate_dfosr: Simulate a dynamic function-on-scalars regression model

In drkowal/dfosr: Dynamic Function-on-Scalars Regression

Description

Simulate data from a dynamic function-on-scalars regression model, allowing for autocorrelated errors and (possibly) dynamic regression coefficients random effects. The predictors are contemporaneously independent but (possibly) autocorrelated.

Usage

simulate_dfosr(T = 200, m = 100, RSNR = 5, K_true = 4, p_0 = 2,
  p_1 = 2, use_dynamic_reg = TRUE, sparse_factors = FALSE,
  use_obs_SV = FALSE, ar1 = 0, prop_missing = 0)

Arguments

T	number of observed curves (i.e., number of time points)
m	total number of observation points (i.e., points along the curve)
RSNR	root signal-to-noise ratio
K_true	rank of the model (i.e., number of basis functions used for the functional data simulations)
p_0	number of true zero regression coefficients
p_1	number of true nonzero regression coefficients
use_dynamic_reg	logical; if TRUE, simulate dynamic regression coefficients; otherwise static
sparse_factors	logical; if TRUE, then for each nonzero predictor j, sample a subset of k=1:K_true factors to be nonzero
use_obs_SV	logical; if TRUE, include stochastic volatility term for the error variance
ar1	AR(1) coefficient for time-correlated predictors
prop_missing	proportion of missing data (between 0 and 1); default is zero

Value

a list containing the following:

• Y: the simulated T x m functional data matrix

• X: the simulated T x p design matrix

• tau: the m-dimensional vector of observation points

• Y_true: the true T x m functional data matrix (w/o noise)

• alpha_tilde_true the true T x p x m array of regression coefficient functions

• alpha_arr_true the true T x p x K_true array of (dynamic) regression coefficient factors

• Beta_true the true T x K_true matrix of factors

• F_true the true m x K_true matrix of basis (loading curve) functions

• sigma_true the true observation error standard deviation

Note

The basis functions (or loading curves) are orthonormalized polynomials, so large values of K_true are not recommended.

Examples

# Example: simulate DFOSR
sim_data = simulate_dfosr()
Y = sim_data$Y; X = sim_data$X; tau = sim_data$tau

drkowal/dfosr documentation built on May 7, 2020, 3:09 p.m.