make_data: Make a synthetic dataset

In spcorum/tsmvrdata: Datasets for tsmvr

Description

Creates a random synthetic dataset for sparse multivariate regression according to the model:

Y = BX + E,

where X is the design matrix, B is the regressor matrix, Y is the response matrix, and E is the matrix error term.

Usage

make_data(n, p, q, b1 = 0.2, b2 = 0.5, sigma = 1, rho_x = 0.6,
  type = "AR1", rho_err = 0.7, h = 0.9, n_edge = 1, shift = 1,
  power = 1, zero_appeal = 1, g = 4, diag_val = 1,
  edge_val = 0.3, reps = 1, seed = NULL)

Arguments

n	number of observations (positive integer)
p	number of regressor features (positive integer)
q	number of responses (positive integer)
b1	Bernoulli parameter for controlling regressor matrix sparsity (positive integer)
b2	another Bernoulli parameter for controlling regressor matrix sparsity (positive integer)
sigma	scale of error term (positive numeric)
rho_x	autoregression parameter for design matrix covariance matrix (0 < rho < 1)
type	type of covariance matrix (string: 'AR1', 'FGN', or 'SFN')
rho_err	autoregression parameter for AR(1) covariance matrix (0 < rho < 1)
h	Hurst parameter for FGN covariance matrix (0 < h < 1)
n_edge	Barabasi algorithm number of edges per step for SFN covariance matrix (positive integer)
shift	eigenvalue shift parameter for SFN covariance matrix (shift > 0) (ensures matrix is PSD) )
power	scaling power for SFN covariance matrix (positive numeric)
zero_appeal	Barabasi algorithm baseline attractiveness for SFN covariance matrix (positive numeric)
g	number of hub nodes for HUB graph precision matrix (positive integer-valued numeric less than q)
diag_val	values of diagonal entries HUB graph precision matrix (non-negative numeric)
edge_val	values of HUB graph network edges
reps	number of randomly drawn datasets to return (positive integer)
seed	seed for pseudo-random number generator

Value

Returns a list of length reps. Each entry is itself a comprising a synthetic sparse multivariate dataset: a regressor matrix B, a design matrix X, and a response matrix Y. B has an expected sparsity of b1 x b2.

See also regressor_matrix, covariance_matrix, and tsmvr_solve.

References

\insertRef

MRCEtsmvr

spcorum/tsmvrdata documentation built on May 6, 2019, 11:17 a.m.