ecdfHT (version 0.1.1)

ecdfHT-package: ecdfHT: A package to plot an empirical cdf for heavy tailed data

Description

The ecdfHT package computes and plot a transformed empirical cdf for data. This is useful because a standard empirical cdf (ecdf) gives little information about the tails of the data when there are extreme values.

Details

The transform is nonparametric: linear in the middle of the data and matched to a log-log transform on the tails, where the tail regions are determined by quantiles. If the data has power law behavior on the tails, the plot is linear on those tails, so this plot can be used as a graphical diagnostic to determine if a data set is heavy tailed.

In addition, there are functions to

• annotate the plot, add custom axes and grid lines
• overlay proposed models on the plot
• fit the tails using linear regression on the transformed tails
• combine the empirical cdf in the middle and the above fit on the tails to get a semi-parametric fit to the data
• compute cdf, pdf, quantiles, and simulate from the semi-parametric fit
• some multivariate plots that look at tail behavior of multiple components and some idea of the dependence.

I will try to fix the code if you provide a simple demonstration of a bug. Polite suggestions for improvements will be considered if there is time available.

See Also

ecdfHT for a basic plot, ecdfHT.draw for annotations and additions to a basic plot, ecdfHT.fit to fit a semi-parametric model to the data, pecdfHT to compute the cdf, pdf, quantiles and simulate from a semi-parametric model, ecdfHT.multivar for multivariate generalizations.