PIR: Parametric Inverse Regression
In abnormally-distributed/cvreg: Cross Validation and Robust Estimation Utilities

Description Usage Arguments Value References

Like sliced inverse regression, parametric inverse regression (PIR) is a method used to compute the effective dimension reduction subspace by finding a set of sufficient predictors that contain all of the information in the model matrix X about the outcome Y (Bura, 1997; Bura & Cook, 2001). PIR was introduced to address the occassional problem with SIR where choosing a different number of slices can yield a different number of sufficient predictors, and hence can be an unreliable estimator of the rank of the effective dimension reduction subspace. Instead of slicing the vector Y into intervals, PIR fits inverse regressions of the form E[X | f(y)] , where f(y) is an m-degree polynomial representation of y. This implementation offers the usual orthogonal polynomials, as well as natural splines and basis splines.

1	PIR(formula, data, rank = "all", df = 3, sm = c("cr", "ns", "bs", "poly"))

`formula`	a model formula
`data`	a data frame
`rank`	the desired number of sufficient predictors to return. the default is "all".
`df`	the tuning parameter used for the parametric inverse regression. defaults to 3.
`sm`	the smoother function to be used. one of cubic regression splines ("cr", the default), basis splines ("bs"), or natural splines ("ns"), or orthogonal polynomials ("poly").

an sdr object

Bura, E. (1997) Dimension reduction via parametric inverse regression. L1-statistical procedures and related topics: Papers from the 3rd International Conference on L1-Norm and Related Methods held in Neuchâtel, 215-228. doi:10.1214/lnms/1215454139

Bura, E., Cook, R.D. (2001) Estimating the structural dimension of regressions via parametric inverse. Journal of the Royal Statistical Society, Series B, 63:393-410.

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