lmXtra: Linear Regression with Non-Gaussian Continuous Distributions
In abnormally-distributed/cvreg: Cross Validation and Robust Estimation Utilities

Description Usage Arguments Details Value

This function contains several families that can be used to model skewed and/or heavy tailed residuals. See details for more.

lmXtra(
  formula,
  data,
  family = c("student", "slash", "shash", "powexp", "skewnorm"),
  nu = "auto",
  alpha = "auto"
)

`formula`	model formula
`data`	data frame
`nu`	the kurtosis parameter. a fixed value can be entered, or if left as the default "auto", it will be estimated.
`alpha`	the skew parameter. a fixed value can be entered, or if left as the default "auto", it will be estimated.

Student's T distribution is characterized by having heavier tails as the degrees of freedom parameter ν tends towards 1 (which defines the cauchy distribution). Student's T distribution only has a defined first and second moment for nu >= 3. Hence, it is capable of modeling outliers through parametric means, and offers an alternative to semi-parametric regression estimators.

The Slash distribution is useful for modeling outliers because it has heavier tails than a normal distribution, but it is not as pathological as the Cauchy distribution. It is defined only by the scale and location, and hence does not require any additional parameters like Student's T.

The Sinh-Arcsinh (shash) distribution is capable of modeling both skewness and kurtosis in addition to the location and scale. This distribution is therefore useful for modeling excess skewness or heavy-tailedness of the residuals. It is functionally similar to a fusion of power exponential type I and type II distribution
.

The skew normal distribution is a continuous probability distribution that modifies the normal distribution to introduce skewness. It is useful for modeling skewed residuals, but it is not resistant to regression outliers.

a "roblm" object

abnormally-distributed/cvreg documentation built on May 3, 2020, 3:45 p.m.