A drawback and an improvement of the classical Weibull probability plot

The classical Weibull Probability Paper (WPP) plot has been widely used to identify a model for fitting a given dataset. It is based on a match between the WPP plots of the model and data in shape. This paper carries out an analysis for the Weibull transformations that create the WPP plot and shows that the shape of the WPP plot of the data randomly generated from a distribution model can be significantly different from the shape of the WPP plot of the model due to the high non-linearity of the Weibull transformations. As such, choosing model based on the shape of the WPP plot of data can be unreliable. A cdf-based weighted least squares method is proposed to improve the parameter estimation accuracy; and an improved WPP plot is suggested to avoid the drawback of the classical WPP plot. The appropriateness and usefulness of the proposed estimation method and probability plot are illustrated by simulation and real-world examples.