tolfind.boxcox {boxcoxmix}

Grid search over tol for NPPML estimation of random effect and variance component models

Description

A grid search over the parameter tol, to set the initial values of the EM algorithm.

Usage

tolfind.boxcox(
  formula,
  groups = 1,
  data,
  K = 3,
  lambda = 1,
  EMdev.change = 1e-04,
  plot.opt = 2,
  s = 15,
  steps = 500,
  find.in.range = c(0, 1.5),
  start = "gq",
  verbose = FALSE,
  noformat = FALSE,
  ...
)

Arguments

formula

a formula describing the transformed response and the fixed effect model (e.g. y ~ x).

groups

the random effects. To fit overdispersion models , set groups = 1.

data

a data frame containing variables used in the fixed and random effect models.

K

the number of mass points.

lambda

a transformation parameter, setting lambda=1 means 'no transformation'.

EMdev.change

a small scalar, with default 0.0001, used to determine when to stop EM algorithm.

plot.opt

Set plot.opt=2, to plot the EM trajectories and the development of the disparity over iteration number. And plot.opt=0, for none of them.

s

number of points in the grid search of tol.

steps

maximum number of iterations for the EM algorithm.

find.in.range

search in a range of tol, with default (0,1.5) in step of 0.1 .

start

a description of the initial values to be used in the fitted model, Quantile-based version "quantile" or Gaussian Quadrature "gq" can be set.

verbose

If set to FALSE, no printed output on progress.

noformat

Set noformat = TRUE, to change the formatting of the plots.

...

extra arguments will be ignored.

Details

A grid search over tol can be performed using tolfind.boxcox() function, which works for np.boxcoxmix() to find the optimal solution.

Value

List with class boxcoxmix containing:

MinDisparity

the minimum disparity found.

Mintol

the value of tol corresponding to MinDisparity.

AllDisparities

a vector containing all disparities calculated on the grid.

Alltol

list of tol values used in the grid.

AllEMconverged

1 is TRUE, means the EM algorithm converged.

aic

the Akaike information criterion of the fitted regression model.

bic

the Bayesian information criterion of the fitted regression model.

Author(s)

Amani Almohaimeed and Jochen Einbeck

See Also

np.boxcoxmix.

Examples

# The Pennsylvanian Hospital Stay Data
data(hosp, package = "npmlreg")
test1 <- tolfind.boxcox(duration ~ age , data = hosp, K = 2, lambda = 0, 
           find.in.range = c(0, 2), s = 10,  start = "gq")
# Minimal Disparity: 137.8368 at tol= 2 
# Minimal Disparity with EM converged: 137.8368 at tol= 2

# Effect of Phenylbiguanide on Blood Pressure
data(PBG, package = "nlme")
test2 <- tolfind.boxcox(deltaBP ~ dose , groups = PBG$Rabbit, find.in.range = c(0, 2),
    data = PBG, K = 2, lambda = -1, s = 15,  start = "quantile", plot.opt = 0)
test2$Mintol
# [1] 1.6
test2$MinDisparity
# [1] 449.5876