| Title: | Generalized Fiducial Inference for Ultrahigh-Dimensional Regression |
| Version: | 1.0.0 |
| Description: | Variable selection for ultrahigh-dimensional ("large p small n") linear Gaussian models using a fiducial framework allowing to draw inference on the parameters. Reference: Lai, Hannig & Lee (2015) <doi:10.1080/01621459.2014.931237>. |
| License: | GPL-3 |
| Encoding: | UTF-8 |
| LazyData: | true |
| RoxygenNote: | 7.1.1 |
| Depends: | R (≥ 3.5.0) |
| Imports: | SIS, stats, utils, lazyeval, mvtnorm |
| URL: | https://github.com/stla/gfiUltra |
| BugReports: | https://github.com/stla/gfiUltra/issues |
| NeedsCompilation: | no |
| Packaged: | 2020-12-07 11:47:55 UTC; stla |
| Author: | Stéphane Laurent [aut, cre] |
| Maintainer: | Stéphane Laurent <laurent_step@outlook.fr> |
| Repository: | CRAN |
| Date/Publication: | 2020-12-09 09:10:02 UTC |
Fiducial confidence intervals for ultrahigh-dimensional regression
Description
Fiducial confidence intervals of the selected parameters of a ultrahigh-dimensional regression.
Usage
gfiConfInt(gfi, conf = 0.95)
Arguments
gfi |
an output of |
conf |
confidence level |
Value
The confidence intervals in a matrix.
See Also
Fiducial estimates for ultrahigh-dimensional regression
Description
Fiducial estimates of the selected parameters of a ultrahigh-dimensional regression.
Usage
gfiEstimates(gfi)
Arguments
gfi |
an output of |
Value
The estimates in a matrix.
See Also
Generalized fiducial inference for ultrahigh-dimensional regression
Description
Generates the fiducial simulations of the parameters of a "large p - small n" regression model and returns the selected models with their probability.
Usage
gfiUltra(formula, data, nsims = 1000L, verbose = FALSE, gamma = 1, ...)
Arguments
formula |
a formula describing the model |
data |
dataframe in which to search the variables of the model |
nsims |
number of fiducial simulations |
verbose |
whether to print the messages generated by the screening procedure |
gamma |
tuning parameter; for expert usage only |
... |
named arguments passed to |
Value
A list with two elements: the fiducial simulations in a matrix
(fidSims) and a vector giving the probabilities of the selected
models (models).
References
Randy C. S. Lai, Jan Hannig & Thomas C. M. Lee. Generalized Fiducial Inference for Ultrahigh-Dimensional Regression. Journal of the American Statistical Association, Volume 110, 2015 - Issue 510, 760-772. <doi:10.1080/01621459.2014.931237>
Examples
# data ####
set.seed(666L)
n <- 300L
p <- 1000L
X <- matrix(rnorm(n * p), nrow = n, ncol = p)
colnames(X) <- paste0("x", 1L:p)
beta <- c(4, 5, 6, 7, 8)
y <- X[, 1L:5L] %*% beta + rnorm(n, sd = 0.9)
dat <- cbind(y, as.data.frame(X))
# fiducial simulations ####
gfi <- gfiUltra(y ~ ., data = dat, nsims = 10000L)
# selected models
gfi$models
# fiducial confidence intervals
gfiConfInt(gfi)
# fiducial estimates
gfiEstimates(gfi)