Plotting positions for fitting distributions and extreme value analysis

Abstract

Random data are often examined on plotting paper to determine appropriate probability distributions and distribution parameters for extreme-value analyses and other applications. Engineers are not always aware of criteria that have been developed and published for selecting the type of probability paper, nor of the impact that the selection can have on results. Some of the advice available in the literature is inaccurate, so it is important that the basis for using the correct method be clearly demonstrated. In the present paper, it is confirmed using straightforward simulation techniques that use of plotting position methods specific to the distribution being considered is appropriate and that use of other plotting position methods can give inaccurate results.