getLimit {extremevalues} | R Documentation |
Determine outlier limit. These functions are called by the wrapperfunction getOutlierLimit
getExponentialLimit(y, p, N, rho) getLognormalLimit(y, p, N, rho) getParetoLimit(y, p, N, rho) getWeibullLimit(y, p, N, rho) getNormalLimit(y, p, N, rho)
y |
Vector of one-dimensional nonnegative data |
p |
Corresponding quantile values |
N |
Number of observations |
rho |
Limiting expexted value |
The functions fit a model cdf to the observed y and p and returns the y-value above which less than rho values are expected, given N observations. See getOutlierLimit for a complete explanation.
The function returns a list with the following entries:
limit |
The y-value above which less then rho observations are expected |
R2 |
R-squared value for the fit |
nFit |
Number of values used in fit (length(y)) |
lamda |
(exponential only) Estimated location (and spread) parameter for f(y)=λexp(-λ y) |
mu |
(lognormal only) Estimated {sf E}(ln(y)) for lognormal distribution |
sigma |
(lognormal only) Estimated Var(ln(y)) for lognormal distribution |
ym |
(pareto only) Estimated location parameter (mode) for pareto distribution |
alpha |
(pareto only) Estimated spread parameter for pareto distribution |
k |
(weibull only) estimated power parameter k for weibull distribution |
lambda |
(weibull only) estimated scaling parameter λ for weibull distribution |
Mark van der Loo, see www.markvanderloo.eu
An outlier detection method for economic data, M.P.J. van der Loo, Submitted to The Journal of Official Statistics (November 2009)
y <- 10^rnorm(50); p <- seq(1,50)/50; L <- getExponentialLimit(y[10:48],p[10:48],50,0.5);