The Identification of “Unusual” Health-Care Providers From a Hierarchical Model

It has become common to adopt a hierarchical model structure when comparing the performance of multiple health-care providers. This structure allows some variation in such measures, beyond that explained by sampling variation, to be “normal,” in recognition of the fact that risk-adjustment is never perfect. The shrinkage estimates arising from such a model structure also have appealing properties.

It is not immediately clear, however, how “unusual” providers, that is, any with particularly high or low rates, can be identified based on such a model. Given that some variation in underlying rates is assumed to be the norm, we argue that it is not generally appropriate to identify a provider as interesting based only on evidence of it lying above or below the population mean. We note with concern, however, that this practice is not uncommon.

We examine in detail three possible strategies for identifying unusual providers, carefully distinguishing between statistical “outliers” and “extremes.” A two-level normal model is used for mathematical simplicity, but we note that much of the discussion also applies to alternative data structures. Further, we emphasize throughout that each approach can be viewed as resulting from a Bayesian or a classical perspective. Three worked examples provide additional insight.     

The “wrong skewness” problem in stochastic frontier models: A new approach

Stochastic frontier models are widely used to measure, e.g., technical efficiencies of firms. The classical stochastic frontier model often suffers from the empirical artefact that the residuals of the production function may have a positive skewness, whereas a negative one is expected under the model, which leads to estimated full efficiencies of all firms. We propose a new approach to the problem by generalizing the distribution used for the inefficiency variable. This generalized stochastic frontier model allows the sample data to have the wrong skewness while estimating well-defined and nondegenerate efficiency measures. We discuss the statistical properties of the model, and we discuss a test for the symmetry of the error term (no inefficiency). We provide a simulation study to show that our model delivers estimators of efficiency with smaller bias than those of the classical model even if the population skewness has the correct sign. Finally, we apply the model to data of the U.S. textile industry for 1958–2005 and show that for a number of years our model suggests technical efficiencies well below the frontier while the classical one estimates no inefficiency in those years.     

The Self‐Similar and Multifractal Nature of a Network Traffic Model

Abstract

We look at a family of models for Internet traffic with increasing input rates and consider approximation models which exhibit self‐similarity at large time scales and multifractality at small time scales. Depending on whether the input rate is fast or slow, the total cumulative input traffic can be approximated by a self‐similar stable Lévy motion or a self‐similar Gaussian process. The stable Lévy limit does not depend on the behavior of the individual transmission schedules but the Gaussian limit does. Also, the models and their approximations show multifractal behavior at small time scales.     

An Empirical Investigation of the “Dynamic McFadden” Model of Purchase Timing and Brand Choice: Implications for Market Structure

Purchase timing and brand-choice decisions of households are jointly investigated using the “dynamic McFadden” model of Heckman and Singer. The hazard of brand purchase is decomposed into the category purchase hazard and the probability of brand choice conditional on category purchase. The former is modeled using the hazard-function approach and the latter using a logit model. Unobserved heterogeneity in brand preferences, marketing effects, and baseline hazard parameters is accounted for in the empirical analysis. The distribution of preference heterogeneity identifies the locations of brands in multiattribute perceptual space and the distribution of attribute importance weights across households.     

A tale of two countries: The Craig–Sakamoto–Matusita theorem

This paper has a long history. Junjiro Ogawa and I have been friends and collaborators for many years. We started this joint paper in the early 1990s and prepared a draft in September 1995. However, we were trying to improve and simplify some of the proofs, and we revised the paper in January 1998. But circumstances were such that we did not submit the paper, but planned to find further simplifications. Junjiro's untimely death in 2003 brought it forth again. Junjiro always liked this topic, and I am pleased to submit this joint paper in his memory. This paper contains both an historical summary as well as several new results. (Ingram Olkin)     

Discussion of “The power of monitoring: how to make the most of a contaminated multivariate sample”

We contribute to the discussion of an article where Andrea Cerioli, Marco Riani, Anthony Atkinson and Aldo Corbellini review the advantages of analyzing multivariate data by monitoring how the estimated model parameters change as the estimation parameters vary. The focus is on robust methods and their sensitivity to the nominal efficiency and breakdown point. In congratulating with the authors for the clear and stimulating exposition, we contribute to its discussion with an overview of what we experienced in applying the monitoring in our application domain.     

A Comparison of Survival Function Estimators in the Koziol–Green Model

In this paper, three competing survival function estimators are compared under the assumptions of the so-called Koziol– Green model, which is a simple model of informative random censoring. It is shown that the model specific estimators of Ebrahimi and Abdushukurov, Cheng, and Lin are asymptotically equivalent. Further, exact expressions for the (noncentral) moments of these estimators are given, and their biases are analytically compared with the bias of the familiar Kaplan–Meier estimator. Finally, MSE comparisons of the three estimators are given for some selected rates of censoring.     

The Effect of Estimation of Unknown Degrees of Freedom on Estimation of Remaining Parameters in Spanos’ Conditional t Heteroskedastic Model

This article investigates the effect of estimation of unknown degrees of freedom on efficient estimation of remaining parameters in Spanos’ conditional t heteroskedastic model. We compare by simulation three maximum likelihood estimators (MLEs) of the remaining parameters in the model: the MLE of the remaining parameters when all the parameters are estimated by the MLE, when the degrees of freedom is estimated by a method of moments estimator, and when the degrees of freedom is erroneously specified. The latter two methods are found to perform poorly compared to the former method for the inference of variance parameters in the model. Thus, efficient estimation of degrees of freedom by the MLE is important to estimate efficiently the remaining variance parameters.     

Comments on “The power of monitoring: how to make the most of a contaminated multivariate sample”

These are comments on the invited paper “The power of monitoring: How to make the most of a contaminated multivariate sample” by Andrea Cerioli, Marco Riani, Anthony Atkinson and Aldo Corbellini.