| Abstract: | Graphical methods are useful for examining residuals after a model is fitted to data. In particular, probability plots can be used to identify peculiar observations and to assess whether a proposed distribution and regression relationship adequately fit the data. A simple extension of probability plotting methods is presented for regression problems where data on the dependent variable are censored on the right. Such data arise, for example, in life tests where the data are analyzed before all test units have failed. The methods are illustrated with censored data from an accelerated life test. |
