The limiting distribution of the maximal deviation of a density estimate and a hazard rate estimate (Q1074975)

This paper presents the large sample theory for the maximal deviation of a density estimator introduced by the second author [Commun. Stat. 2, 493-506 (1973; Zbl 0283.62040)]. This estimator is based on the random spacings of the order statistics of observations. The normalized maximal deviation of this estimator over a compact interval is shown to have limiting distribution related to the maximum of a Gaussian process. The use of this result to form uniform confidence bands for the density function and its relation to goodness-of-fit is discussed. As a natural extension, an hazard rate estimator is proposed based on this density estimator. Similar limiting distributions are also established accordingly. In the context of random censoring, an hazard estimator is constructed based on similar ideas and is studied in a following-up paper.