Visualisation and validation of weibull models; A new diagnostic methodology

This paper contends that analysis based on the traditional Weibull Plot (WP), using probability scales, has limited scope for providing appropriate visualization and validation against a backdrop of endemic data heterogeneity and a frequent need to analyze data that is ill-defined. An extension to the WP is proposed, namely the Aggregate Cumulative Hazard (ACH) Plot, used in conjunction with Maximum Likelihood Estimation (MLE). This has universal flexibility across all types of data. It is intuitive, allowing for easier interpretation by a wider body of stakeholders. It can be specifically used to identify heterogeneity by examining the influence of many covariates. As such it has powerful diagnostic capability, allowing statistical support of root cause analysis. It can be used in conjunction with simple two parameter Weibull models but also with more complex models ranging from competing risks, mixtures and the whole range of accelerated life models including proportional hazards. In short, it is proposed as a one stop shop for supporting the validation and guidance of any type of Weibull-based analysis.