Search

This article describes prediction methods for the number of future events from a population of units associated with an on-going time-to-event process. Examples include the prediction of warranty returns and the prediction of the number of future product failures that could cause serious threats to property or life. Important decisions such as whether a product recall should be mandated are often based on such predictions. Data, generally right-censored (and sometimes left truncated and right-censored), are used to estimate the parameters of a time-to-event distribution. This distribution can then be used to predict the number of events over future periods of time. Such predictions are sometimes called within-sample predictions and differ from other prediction problems considered in most of the prediction literature. This article shows that the plug-in (also known as estimative or naive) prediction method is not asymptotically correct (i.e., for large amounts of data, the coverage probability always fails to converge to the nominal confidence level). However, a commonly used prediction calibration method is shown to be asymptotically correct for within-sample predictions, and two alternative predictive-distribution-based methods that perform better than the calibration method are presented and justified. for this article are available online. [ABSTRACT FROM AUTHOR]

Random forests are among the most popular machine learning techniques for prediction problems. When using random forests to predict a quantitative response, an important but often overlooked challenge is the determination of prediction intervals that will contain an unobserved response value with a specified probability. We propose new random forest prediction intervals that are based on the empirical distribution of out-of-bag prediction errors. These intervals can be obtained as a by-product of a single random forest. Under regularity conditions, we prove that the proposed intervals have asymptotically correct coverage rates. Simulation studies and analysis of 60 real datasets are used to compare the finite-sample properties of the proposed intervals with quantile regression forests and recently proposed split conformal intervals. The results indicate that intervals constructed with our proposed method tend to be narrower than those of competing methods while still maintaining marginal coverage rates approximately equal to nominal levels. [ABSTRACT FROM AUTHOR]

A probability model exhibits instability if small changes in a data outcome result in large and, often unanticipated, changes in probability. This instability is a property of the probability model, given by a distributional form and a given configuration of parameters. For correlated data structures found in several application areas, there is increasing interest in identifying such sensitivity in model probability structure. We consider the problem of quantifying instability for general probability models defined on sequences of observations, where each sequence of length |$N$| has a finite number of possible values that can be taken at each point. A sequence of probability models, indexed by |$N$|⁠ , and an associated parameter sequence result to accommodate data of expanding dimension. Model instability is formally shown to occur when a certain log probability ratio under such models grows faster than |$N$|⁠. In this case, a one component change in the data sequence can shift probability by orders of magnitude. Also, as instability becomes more extreme, the resulting probability models are shown to tend to degeneracy, placing all their probability on potentially small portions of the sample space. These results on instability apply to large classes of models commonly used in random graphs, network analysis and machine learning contexts. [ABSTRACT FROM AUTHOR]

For spatial and network data, we consider models formed from a Markov random field (MRF) structure and the specification of a conditional distribution for each observation. Fast simulation from such MRF models is often an important consideration, particularly when repeated generation of large numbers of datasets is required. However, a standard Gibbs strategy for simulating from MRF models involves single-site updates, performed with the conditional univariate distribution of each observation in a sequential manner, whereby a complete Gibbs iteration may become computationally involved even for moderate samples. As an alternative, we describe a general way to simulate from MRF models using Gibbs sampling with "concliques" (i.e., groups of nonneighboring observations). Compared to standard Gibbs sampling, this simulation scheme can be much faster by reducing Gibbs steps and independently updating all observations per conclique at once. The speed improvement depends on the number of concliques relative to the sample size for simulation, and order-of-magnitude speed increases are possible with many MRF models (e.g., having appropriately bounded neighborhoods). We detail the simulation method, establish its validity, and assess its computational performance through numerical studies, where speed advantages are shown for several spatial and network examples. for this article are available online. [ABSTRACT FROM AUTHOR]