simulateYX2: Generating Random Variates for Simulation Studies with...

In ipeglim: Imprecise Inferential Framework on Statistical Reasoning

Description

simulateYX2 generates random variates of a response variable under the given conditions of mean and variance-covariance structure of matrices of predictors and their coefficients.

Usage

  simulateYX2(N, Ystr = list(ymean, ysigma), param1,
    param2,
    Xstr = list(type = c("mvnorm", "mvbion", "mixed", "multinomial"), Xmean, XR, Xsd, Xmat = NULL),
    shape = NULL, ztrunc, xreg, seed = NULL, ...)

Arguments

N	a number of random variates
Xstr	a structure of covariates X
ztrunc	FALSE by default
xreg	FALSE by default
seed	NULL a random seed
shape	a shape parameter of gamma distribution
...	other arguments
Ystr	structure of responses
param1	a vector of regression parameters 1
param2	a vector of regression parameters 2

Details

The mean parameter of a response variable is a function of linear predictor by a link function in the context of generalized linear models.

Log and logit links are supported.

Default to Xstr is an independent multivariate normal distribution with mean Xmean, correlation R, and standard deviation Xsd.

Value

a list with components

n	the number of observations which are actually realized
y	the n-by-1 vector of response generated
X	the n-by-p-1 matrix (an intercept is excluded)
yX	the n-by-p data frame which includes y and X

TODO

• A component with coefficients of predictors is not implemented yet.

• Multinomal type of predictor is not incorporated yet.

FIXME

• No bugs are reported yet.

Note

Default to a matrix decomposition method is eigen. Other options are chol and svd.

Currently, the link logit is not supported.

Author(s)

Chel Hee Lee <[e-mail]>

See Also

rmvnorm

Examples

# do not run

ipeglim documentation built on May 2, 2019, 4:31 p.m.