Graphical methods for plotting and evaluating Weibull distributed data

Aiming at making Weibull plots, three types of plotting positions are reviewed: median probability, mean probability and mean plotting position. It is important whether the (log)data are regarded as median or as mean values. Median type plotting gives a biased plot on the average and, if used to identify outliers, can be somewhat misleading at the low end of the curve. In contrast, the mean type gives unbiased estimators by linear regression estimation, thus granting the best Weibull plot on the average. For use of Weibull paper in combination with a calculator, adequate formulas approximating probabilities are discussed. A somewhat improved formula for expected plotting position is given.<>