mvtnorm: qmvnorm – R documentation

Become an expert in R — Interactive courses, Cheat Sheets, certificates and more!

qmvnorm

Quantiles of the Multivariate Normal Distribution

Description

Computes the equicoordinate quantile function of the multivariate normal distribution for arbitrary correlation matrices based on inversion of pmvnorm, using a stochastic root finding algorithm described in Bornkamp (2018).

Usage

qmvnorm(p, interval = NULL, tail = c("lower.tail", 
        "upper.tail", "both.tails"), mean = 0, corr = NULL, 
        sigma = NULL, algorithm = GenzBretz(),
        ptol = 0.001, maxiter = 500, trace = FALSE, ...)

Arguments

`p`	probability.
`interval`	optional, a vector containing the end-points of the interval to be searched. Does not need to contain the true quantile, just used as starting values by the root-finder. If equal to NULL a guess is used.
`tail`	specifies which quantiles should be computed. `lower.tail` gives the quantile x for which P[X ≤ x] = p, `upper.tail` gives x with P[X > x] = p and `both.tails` leads to x with P[-x ≤ X ≤ x] = p.
`mean`	the mean vector of length n.
`corr`	the correlation matrix of dimension n.
`sigma`	the covariance matrix of dimension n. Either `corr` or `sigma` can be specified. If `sigma` is given, the problem is standardized. If neither `corr` nor `sigma` is given, the identity matrix is used for `sigma`.
`algorithm`	an object of class `GenzBretz`, `Miwa` or `TVPACK` specifying both the algorithm to be used as well as the associated hyper parameters.
`ptol, maxiter, trace`	Parameters passed to the stochastic root-finding algorithm. Iteration stops when the 95% confidence interval for the predicted quantile is inside [p-ptol, p+ptol]. `maxiter` is the maximum number of iterations for the root finding algorithm. `trace` prints the iterations of the root finder.
`...`	additional parameters to be passed to `GenzBretz`.

Details

Only equicoordinate quantiles are computed, i.e., the quantiles in each dimension coincide. The result is seed dependend.

Value

A list with two components: quantile and f.quantile give the location of the quantile and the difference between the distribution function evaluated at the quantile and p.

References

Bornkamp, B. (2018). Calculating quantiles of noisy distribution functions using local linear regressions. Computational Statistics, 33, 487–501.

Examples

qmvnorm(0.95, sigma = diag(2), tail = "both")

mvtnorm

Multivariate Normal and t Distributions

v1.1-1

GPL-2

Authors

Alan Genz [aut], Frank Bretz [aut], Tetsuhisa Miwa [aut], Xuefei Mi [aut], Friedrich Leisch [ctb], Fabian Scheipl [ctb], Bjoern Bornkamp [ctb] (<https://orcid.org/0000-0002-6294-8185>), Martin Maechler [ctb], Torsten Hothorn [aut, cre] (<https://orcid.org/0000-0001-8301-0471>)

Initial release

2020-06-09