poissonMTinitialParallel: Initial Robust Estimates based on MT robust method for... In poissonMT: Robust M-Estimators Based on Transformations for Poisson Model

Description

`poissonMTinitialParallel` is used to provides a robust initial estimate for fit generalized linear models. This is the parallel computing version. The model is specified by the `x` and `y` components.

Usage

 ```1 2 3 4 5``` ```poissonMTinitialParallel(x, y, stage2 = TRUE, alpha = c(0.025, 0.025), tol = 1e-04, cc = 2.3, psi = "bisquare", maxit = 20, zero = sqrt(.Machine\$double.eps), replace.small = 0.5, start = NULL, na.to.zero = TRUE, parallel = c("no", "multicore", "snow"), ncpus = 1, cl = NULL) ```

Arguments

 `x` design matrix of dimension n * p. `y` vector of observations of length `n`. `stage2` logical, the second stage should be performed? `alpha` quantile orders used in the second stage. `tol` convergence tolerance for the parameter vector. `cc` tuning constant c for Tukey's bisquare psi-function. `psi` the name of the `psi` function. At the moment only the `bisquare` is available. `maxit` integer specifying the maximum number of IRWLS iterations. `zero` eigenvalues smaller than `zero` will be considered exactly equal to 0. `replace.small` all the observations `y` smaller than `replace.small` are replaced by `replace.small` value. `start` eventual starting values, as a reference, for the parameters in the linear predictor. `na.to.zero` logical, should the eventual `NA` in the coefficients be replaced by `0`? `parallel` The type of parallel operation to be used. By default (`none`) no parallel is used. `ncpus` integer: number of processes to be used in parallel operation. Typically one would chose this to the number of available CPUs. `cl` An optional `parallel` or `snow` cluster for use if `parallel = "snow"`. If not supplied, a cluster on the local machine is created for the duration of the `poissonMTinitialParallel` call.

Details

This function is the same as function `poissonMTinitial`, however it can takes advantage of parallel computing.

Value

A list with the following components

 `coefficients1` initial value proposed at the end of the first stage. `obj1` value of the MT objective function at `coefficients1`. `coefficients2a` initial value proposed at the end of the first part od the second stage. `obj2a` value of the MT objective function at `coefficients2a`. `coefficients2b` initial value proposed at the end of the second part od the second stage. `obj2b` value of the MT objective function at `coefficients2b`. `coefficients` initial value proposed. `obj` value of the MT objective function at `coefficients`.

Author(s)

Claudio Agostinelli, Marina Valdora and Victor J. Yohai

References

C. Agostinelli, M. Valdora and V.J Yohai (2018) Initial Robust Estimation in Generalized Linear Models with a Large Number of Covariates. Submitted.

M. Valdora and V.J. Yohai (2014) Robust estimators for generalized linear models. Journal of Statistical Planning and Inference, 146, 31-48.

`poissonMTinitial`, `poissonMT` and `poissonL2T`
 ```1 2 3 4 5``` ``` data(epilepsy) x <- model.matrix( ~ Age10 + Base4*Trt, data=epilepsy) poissonMTsetwd(tempdir()) start <- poissonMTinitialParallel(x=x, y=epilepsy\$Ysum)\$coefficients start ```