influence: Leverage and Local Influence for the Unit Gamma Regression

In rdmatheus/ugrpl: Unit Gamma Regression with Parametric Link Functions for Modeling Rates and Proportions

Leverage and Local Influence for the Unit Gamma Regression

Description

Leverage and Local Influence for the Unit Gamma Regression

Usage

influence(
  object,
  scheme = c("case-weight", "mean", "dispersion", "both"),
  covariate,
  plot = TRUE,
  ask = prod(graphics::par("mfcol")) < length(which) && grDevices::dev.interactive(),
  ...
)

Arguments

object	an "ugrpl" object.
scheme	character; it specifies the perturbation scheme. Currently, the following perturbation schemes are available: case-weight perturbation ("case-weight"); perturbation of the mean covariates ("mean"); perturbation of the dispersion covariates ("dispersion"); and simultaneous perturbation of the mean and dispersion covariates ("both"). If the mean and the dispersion submodels share the same covariate, then it is not possible to use scheme = "mean" or scheme = "dispersion".
covariate	character; it specifies the name of the continuous variable used in the covariate perturbation scheme. If scheme = "both", then it must be a vector of size 2, where the first element is the name of the perturbed covariate associated with the mean and the second element is the name of the perturbed covariate associated with the dispersion. If scheme = "mean" or scheme = "dispersion", then the mean and dispersion submodels cannot share the same perturbed covariate.
plot	logical. If TRUE, the graphs of local influence and generalized leverage for each sampled observation are presented.
ask	logical. If TRUE, the user is asked before each plot.
...	further arguments passed plot.

Details

Leverage points are observations that have a disproportionate weight in their fitted value. These points, in general, present a different behavior from the other points in relation to the values of the explanatory variables and can strongly influence the estimates of the regression coefficients. On the other hand, local influence analysis should be considered when there is an interest in investigating the sensitivity of postulated assumptions under small perturbations in the model or data.

The leverage measure considered is based on the work of Wei et al. (1998) and local influence perturbation schemes are commonly considered schemes, see, for instance Cook (1986).

Value

A list containing the following components:

dmax

local influence under the selected perturbation scheme.

total.LI

total local influence.

scheme

name of the considered perturbation scheme

leverage

the diagonal elements of the generalized leverage matrix.

Author(s)

Rodrigo M. R. de Medeiros <[e-mail]>

References

Cook, R. D. (1986). Assessment of local influence. Journal of the Royal Statistical Society B, 48, 133–155.

Wei, B. C., Hu, Y. Q., & Fung, W. K. (1998). Generalized leverage and its applications. Scandinavian Journal of Statistics, 25, 25–37.

rdmatheus/ugrpl documentation built on July 13, 2024, 10:24 p.m.