loggammarob.control: Tuning parameters for loggammarob
In robustloggamma: Robust Estimation of the Generalized log Gamma Model

Description Usage Arguments Author(s) References See Also Examples

View source: R/loggammarob.R

Tuning parameters for 'loggammarob' for all four methods implemented.

loggammarob.control(method="oneWL", tuning.rho=1.547647,
  tuning.psi=6.08, lower=-7, upper=7, n=201, max.it=750,
  refine.tol=1e-6, solve.tol=1e-7, nResample=100, bw=0.3,
  smooth=NULL, raf=c("NED","GKL","PWD","HD","SCHI2"),
  tau=1, subdivisions=1000, lambda.step=TRUE, sigma.step=TRUE,
  step=1, minw=0.04, nexp=1000, reparam=NULL, bootstrap=FALSE,
  bootstrap.lambda=NULL, qthreshold=0.9, nTML=2000, xmax=1e100,
  iter=1, pcut=0.997, compute.rd=FALSE,
  eps.outlier= function(nobs) 0.1 / nobs)

`method`	character. The method to be used. See Details below.
`tuning.rho`	numeric. Tuning constant c1 for the Tau-estimator.
`tuning.psi`	numeric. Tuning constant c2 for the Tau-estimator.
`lower`	numeric. The lower limit for the search grid of the shape parameter.
`upper`	numeric. The upper limit for the search grid of the shape parameter.
`n`	numeric. The number of subdivisions for the search grid of the shape parameter.
`max.it`	numeric. Maximum number of iterations.
`refine.tol`	numeric. Relative convergence tolerance for the fully iterated best candidates.
`solve.tol`	numeric. Relative tolerance for inversion. Hence, this corresponds to `solve.default()`'s `tol`.
`nResample`	integer. Number of re-sampling candidates to be used to find the initial estimator. Currently defaults to 100 which works well in most situations.
`bw`	numeric. Bandwidth used in the Weighted Likelihood steps.
`smooth`	`NULL` or numeric. When not `NULL` the parameter `bw` is set to `smooth` times the square root of the starting value of the scale parameter.
`raf`	character. Residual adjustment function used in the Weighted Likelihood steps `raf="NED"`: Negative Exponential Disparity RAF, `raf="GKL"`: Generalized Kullback-Leibler Divergence Family with parameter `tau` (see below) RAF. `raf="PWD"`: Power Divergence Family with parameter `tau` (see below) RAF. `raf="HD"`: Hellinger Distance RAF, `raf="SCHI2"`: Symmetric Chi-Squared Disparity RAF. Default value is "NED".
`tau`	parameter used when `raf` is equal to "PWD" or "GKL".
`subdivisions`	numeric. Number of subdivisions used in the approximation of the smoothed model density in the Weighted Likelihood steps.
`lambda.step`	logical.
`sigma.step`	logical.
`step`	integer. Number of steps to be performed when `method` is "oneWL" (only implemented for the functions for non censored data).
`minw`	numeric. A scalar in the interval [0,1]. When `method` is "oneWL" the weights smaller than `minw` are set to zero.
`nexp`	integer. When `method` is "oneWL" number of quantile points used in the approximation of the Expected Jacobian matrix.
`reparam`	list. When `method` is "oneWL" a reparametrization is possible for the "sigma" parameter. See function `sqrtloggamma` for an example.
`bootstrap`	logical. To use `loggammarob` in `boot`
`bootstrap.lambda`	numeric. An initial estimates for the shape parameter. To use `loggammarob` in `boot`
`qthreshold`	numeric. A value in (0.5, 1] used for TQtau e TWQtau procedure. It is the quantile order to truncated the data on the right.
`nTML`	numeric. Number of elements to be considered in the grid for finding the cut points of the TML.
`xmax`	numeric. A threshold value for the log likelihood. Used for ML.
`iter`	numeric. Number of iterations to be performed. Only working for TML.
`pcut`	numeric. Fraction to determined the cut points of the TML.
`compute.rd`	logical. Indicating if robust distances (based on the MCD robust covariance estimator `covMcd`) are to be computed for the robust diagnostic plots. This may take some time to finish, particularly for large data sets, and can lead to singularity problems when there are `factor` explanatory variables (with many levels, or levels with "few" observations). Hence, is `FALSE` by default.
`eps.outlier`	limit on the robustness weight below which an observation is considered to be an outlier. Either a `numeric(1)` or a function that takes the number of observations as an argument. Used in `summary.loggammacenslmrob`.

C. Agostinelli, A. Marazzi, V.J. Yohai and A. Randriamiharisoa

C. Agostinelli, A. Marazzi and V.J. Yohai (2015) Robust estimates of the generalized loggamma distribution, Technometrics, Volume 56, Issue 1, 2014. DOI: 10.1080/00401706.2013.818578

C. Agostinelli, A. Marazzi, V.J. Yohai and A. Randriamiharisoa (2016)

Robust Estimation of the Generalized Loggamma Model. The R Package robustloggamma. Journal of Statistical Software. Accepted.

C. Agostinelli, I. Locatelli, A. Marazzi and V.J. Yohai (2015) Robust estimators of accelerated failure time regression with generalized log-gamma errors. Submitted.

loggammarob