Estimation of the probability content in a specified interval using fiducial approach.

Statistical methods for constructing confidence intervals for the probability content in a specified interval are proposed. Exact and approximate solutions based on the fiducial approach are described when the measurements on the variable of interest can be modelled by a location-scale (or log-location-scale) distribution. Methods are described for the normal, Weibull, two-parameter exponential and two-parameter Rayleigh distributions. For each case, the solutions are evaluated for their merits. Three examples, where it is desired to estimate the percentages of engineering products meet the specification limits, are provided to illustrate the methods.

Confidence interval, prediction interval and tolerance limits for a two-parameter Rayleigh distribution.

The problems of interval estimating the parameters and the mean of a two-parameter Rayleigh distribution are considered. We propose pivotal-based methods for constructing confidence intervals for the mean, quantiles, survival probability and for constructing prediction intervals for the mean of a future sample. Pivotal quantities based on the maximum likelihood estimates (MLEs), moment estimates (MEs) and the L-moments estimates (L-MEs) are proposed. Interval estimates based on them are compared via Monte Carlo simulation. Comparison studies indicate that the results based on the MEs and the L-MEs are very similar. The results based on the MLEs are slightly better than those based on the MEs and the L-MEs for small to moderate sample sizes. The methods are illustrated using an example involving lifetime data.