MONITORING PROCESS LOCATION AND DISPERSION SIMULTANEOUSLY USING A CONTROL REGION

Control charts are effective tools for detecting out-of-control conditions of process parameters in manufacturing and service industries. The development of distribution-free control charts is important in statistical process control when the process quality variable follows an unknown or a non-normal distribution. This research thus proposes to use a distribution-free technology to establish a new control region based on the exponentially weighted moving average median statistic and exponentially weighted moving average interquartile range statistic for simultaneously monitoring the process location and dispersion and further sets up a corresponding new control chart. We compute the out-of-control average run length to evaluate out-of-control detection performance of the proposed control region and also compare the proposed control region with some existing location and dispersion control charts. Results show that our proposed chart always exhibits superior detection performance when the shifts in process location and/or dispersion are small or moderate. The new control region is thus recommended. Keywords: control chart, distribution-free, dispersion and location, EWMA, kernel control region, kernel density estimation.

Distribution-free, Risk-controlling Prediction Sets

While improving prediction accuracy has been the focus of machine learning in recent years, this alone does not suffice for reliable decision-making. Deploying learning systems in consequential settings also requires calibrating and communicating the uncertainty of predictions. To convey instance-wise uncertainty for prediction tasks, we show how to generate set-valued predictions from a black-box predictor that controls the expected loss on future test points at a user-specified level. Our approach provides explicit finite-sample guarantees for any dataset by using a holdout set to calibrate the size of the prediction sets. This framework enables simple, distribution-free, rigorous error control for many tasks, and we demonstrate it in five large-scale machine learning problems: (1) classification problems where some mistakes are more costly than others; (2) multi-label classification, where each observation has multiple associated labels; (3) classification problems where the labels have a hierarchical structure; (4) image segmentation, where we wish to predict a set of pixels containing an object of interest; and (5) protein structure prediction. Last, we discuss extensions to uncertainty quantification for ranking, metric learning, and distributionally robust learning.

Joint pricing and ordering decisions for a loss-averse retailer with quantity-oriented reference point effect and demand uncertainty: a distribution-free approach

PurposeThis work examines the joint pricing and ordering (JPO) decisions for a loss-averse retailer with quantity-oriented reference point (RP) effect under demand uncertainty.Design/methodology/approachThe demand is assumed to be uncertain with the mean and variance as the only known information. The prospect theory is used to model the retailer's expected utility. An expected utility maximization model in the distribution-free approach (DFA) is then developed. Using duality theory, the expected utility under the worst-case distribution is transformed into tractable piece-wise functions. To examine the effectiveness of the DFA in coping with the demand uncertainty, a stochastic programming model is developed and its solutions are used as benchmarks.FindingsThe proposed model and solution approach can effectively hedge against the demand uncertainty. The JPO decisions are significantly influenced by the LA coefficient and the reference level. The LA has a stronger influence than the reference level does on the expected utility. An excessive LA is detrimental while an appropriate reference level is beneficial to the retailer.Practical implicationsThe results of this work are applicable to loss-averse retailers with the quantity-oriented RP when making JPO decisions with difficulty in predicting the demands.Originality/valueThe demand is assumed to be uncertain in this work, but a certain demand distribution is usually assumed in the existing literature. The DFA is used to study JPO decisions for the loss-averse retailer with quantity-oriented RP effect under the uncertain demand.