liftone_GLM: Unconstrained lift-one algorithm to find D-optimal...

In CDsampling: 'CDsampling': Constraint Sampling in Paid Research Studies

Unconstrained lift-one algorithm to find D-optimal allocations for GLM

Description

Unconstrained lift-one algorithm to find D-optimal allocations for GLM

Usage

liftone_GLM(
  X,
  W,
  reltol = 1e-05,
  maxit = 500,
  random = TRUE,
  nram = 3,
  w00 = NULL
)

Arguments

X	Model matrix, with nrow = num of design points and ncol = num of parameters
W	Diagonal of W matrix in Fisher information matrix, can be calculated from W_func_GLM in package
reltol	The relative convergence tolerance, default value 1e-5
maxit	The maximum number of iterations, default value 500
random	TRUE or FALSE, if TRUE then the function will run with additional "nram" number of random initial points, default to be TRUE
nram	When random == TRUE, the function will generate nram number of initial points, default is 3
w00	Specified initial design proportion; default to be NULL, this will generate a random initial design

Value

w is the approximate D-optimal design

w0 is the initial design used to get optimal design w

Maximum is the maximized |F| value

itmax is the number of iterations

convergence is TRUE or FALSE, if TRUE means the reported design is converged

Examples

beta = c(0.5, 0.5, 0.5)
X = matrix(data=c(1,-1,-1,1,-1,1,1,1,-1), byrow=TRUE, nrow=3)
W_matrix = W_func_GLM(X=X, beta=beta)
w00 = c(1/6, 1/6, 2/3)
approximate_design = liftone_GLM(X=X, W=W_matrix, reltol=1e-10, maxit=100,
random=FALSE, nram=3, w00=w00)

CDsampling documentation built on Oct. 13, 2024, 9:07 a.m.