Citation
Zainal Abidin, Nahdiya and Adam, Mohd Bakri and Midi, Habshah
(2012)
Hybrid conditional plot of goodnessoffit for Gumbel distribution.
Journal of Quality Measurement and Analysis, 8 (1).
pp. 4353.
ISSN 18235670
Abstract
A Gumbel model is an extreme value model that describes the event of extreme behaviour. The Gumbel model has an exponential tail. Generally, the goodnessoffit for the Gumbel model is evaluated by the graphical form of probability plot (PP) and quantiles plot (QQ). The model fits the observed values if the probability and the quantiles of the hypothetical distribution are linearly plotted against that of the observed values. However, the QQ plot is quite sensitive to the deviation at the tail of the plot, as opposed to the PP plot which is somewhat robust. Thus, distribution of extreme values is likely to deviate from the linear line at the tail of the QQ plot. An alternative approach of plotting the Gumbel model is given, in which the approach is expected to produce the linear plot. The conditional plot and stabilised plot are employed and the performances of both are compared. The plots are transformed into the hybrid plot so that the departures of the hypothetical quantiles values from the observed quantiles values are illustrated. The result shows that the hybrid conditional QQ plot is a better plot of goodnessoffit for Gumbel model.
Download File
Preview 

PDF (Abstract)
Hybrid conditional plot of goodnessoffit for Gumbel distribution.pdf
Download (35kB)
 Preview


Additional Metadata
Actions (login required)

View Item 