predint: Prediction intervals to contain at least k out of m future...

In daniel-gerhard/predIntervals: predIntervals

Description

Calculates parametric and non-parametric prediction intervals to contain at least k out of m future observations given a observed sample

Usage

predint(x, k, m, level=0.95, alternative="two.sided", quantile=NULL, absError=0.001, interval=c(0, 100))
nparpredint(x, k, m, level=0.95, alternative="two.sided")

Arguments

x	A numeric vector with a sample of observations
k	Number of future observations contained in the interval
m	Size of future sample
level	Confidence level
alternative	One of "two.sided", "less", or "greater" to calculate two- or one-sided intervals
quantile	Supply a user-defined quantile for parametric intervals
absError	The maximum absolute error tolerated when calculating the quantile
interval	Root finding interval when calculating the quantile

Value

An object inheriting from class PredInterval, or nparPredInterval

References

Odeh, RE (1990): 2-Sided prediction intervals to contain at least k out of m future observations from a normal distribution. Technometric 32(2): 203-216.

Danziger, L. and Davis, S.A. (1964): Tables of distribution-free tolerance limits. Annals of Mathematical Statistics 35(3):1361-1365.

See Also

precint

Examples

x <- rnorm(50)
predint(x, k=7, m=10)
nparpredint(x, k=7, m=10)

## Not run: 
# for k=m the prediction interval:
predint(x, k=10, m=10, absError=0.0001)

# can also be calculated by
library(mvtnorm)
rho <- matrix(1/(length(x)+1), nrow=10, ncol=10)
diag(rho) <- 1
m <- mean(x) 
std <- sd(x)
quant <- qmvt(0.95, df=length(x)-1, corr=rho, tail="both.tails")$quantile
m + c(-1, 1) * quant * std * sqrt((length(x)+1)/length(x))

## End(Not run)

daniel-gerhard/predIntervals documentation built on May 14, 2019, 3:38 p.m.