Band 27, Heft 1

Abstract. A new two-parameter lifetime distribution with increasing failure rate is introduced. Various properties of the proposed distribution are discussed. The distribution parameters are estimated by the EM algorithm and the asymptotic variances and covariances of these estimators are obtained. The accuracy of maximum likelihood estimation of variances and covariances are studies by simulation. Some experimental results based on real data sets illustrate the results.

Abstract. The absence of a valid and reliable scale to measure e-service quality forced early researchers to make use of some fairly unsatisfactory alternatives. To improve these previous studies, Parasuraman et al. (2005) developed and operationalized E-S-QUAL, a multi-item scale for examining website service quality. This study reviews the literature on E-S-QUAL focusing particularly on the methodology used, new suggestions to be made, and limitations associated with the adoption of the scale. The review shows that the scale is effective in capturing core e-service quality. The dimensions of efficiency, system availability and privacy are represented consistently in the various models, regardless of the type of e-service. In contrast, the dimension of fulfillment appears not to be generic but specific to particular e-service contexts such as websites selling physical goods. This study is the first to undertake an extensive literature review of the development of e-service quality scales and to provide important insights into the application of E-S-QUAL.

Abstract. The construction of a statistical test is investigated which is based only on “reliability” and “precision” as quality criteria. The reliability of a statistical test is quantified in a straightforward way by the probability that the decision of the test is correct. However, the quantification of the precision of a statistical test is not at all evident. Therefore the paper presents and discusses several approaches. Moreover the distinction of “null hypothesis” and “alternative hypothesis” is not necessary any longer.

Abstract. Repairable systems are characterized by alternation of function times and repair times. The repair times are often neglected, and the investigation of the system is based on sequences of failure times and the corresponding counting processes. Nonhomogeneous Poisson processes describe cases of minimal repair. Renewal processes model perfect repairs. More general models for repair types between the cases of minimal and perfect repair are studied by Kijima and Lindqvist. For different repair levels Kijima introduced the virtual age function, and the failure intensity of the system depends on the virtual age. Trend-renewal processes (TRP) introduced by Lindqvist generalize renewal processes of failure time points and lead to general failure intensities. The paper introduces a new model for repairable systems: a combination of TRP and Kijima virtual age models. The generalization contains most models dealt with in literature as special cases. For special process examples, maximum likelihood and Bayes parameter estimation are considered.

Abstract. Distributional properties of logistic order statistics subject to independent exponential one-sided and two-sided shifts are established. Utilizing these properties, we extend several known results and obtain new characterizations of the logistic distribution.