RFPE: Robust final prediction error for a linear model
In harrysouthworth/margarita: Support functions for texmex

Description Usage Arguments Details Value References

Robust final prediction error for a linear model

1	RFPE(x, scale)

`x`	A robust linear model, fit by `margarita::lmr` or `robust::lmRob` or `robustbase::lmrob`, or a list containing multiple such models.
`scale`	The value of the scale to use. If `x` is a list and `scale` is not provided by the user, the function takes the scale estimate to be the one associated with the largest model. Otherwise, no default is available. See details.

The definition of robust final prediction error is given in Section 5.12 of Maronna et al. The function is generic, but so far only methods for objects fit by lmr, lmrob, lmRob and for a list of such models have been implemented. Only bisquare weight functions are considered. Note that the purpose of RFPE is to compare models – for a single model, it's not very meaningful. To allow a valid comparison, the calculation must use the same scale estimate, usually the one taken from the largest model. For this reason, no default scale argument is provided, forcing the user to specify a value.

A number representing the robust final prediction error or, if x is a list, a data.frame giving the final prediction error and other details.

Maronna, R. A, Martin, R. D and Yohai, V. J. (2006) Robust Statistics: Theory and Methods, Wiley

harrysouthworth/margarita documentation built on Aug. 19, 2021, 5 a.m.