Probability plots and order statistics of the standard extreme value distribution (Q2430242): Difference between revisions

The effect of weights based on the variances and covariances in regression analysis when data come from a standard extreme value distribution [see \textit{J.S. White}, Technometrics 11, 373--386 (1969; Zbl 0174.22904)] is studied. It is shown in a probability plot that the unreliable observations influence the estimation of the unknown parameters and that the use of a weighted least-squares estimator (WLSE) improves the least-squares line [see \textit{D.R. Cook} and \textit{S. Weisberg}, An introduction to regression gaphics. NY: Wiley (1994; Zbl 0925.62287); \textit{R.J. Carroll} and \textit{D. Ruppert}, Transformation and weighting in regression. NY etc.: Chapman and Hall (1988; Zbl 0666.62062); \textit{N. R. Draper} and \textit{H. Smith}, Applied regression analysis. 3rd ed. NY: Wiley (1998; Zbl 0895.62073)]. A test statistic based on WLSE is proposed and the percentage points of the test statistic are presented.