Fast retraction of evidence in a probabilistic expert system

We present and justify a propagation algorithm to facilitate the simultaneous calculation, for every node in a probabilistic exper system of the distribution of the associated random quantity, conditional on all the evidence obtained about the remaining nodes.     

To be or not to be involved: a questionnaire-experimental view on Harsanyi’s utilitarian ethics

According to standard theory founded on Harsanyi (J Polit Econ 61:434–435, 1953; 63:309–321, 1955) a social welfare function can be appropriately based on the individual’s approach to choice under uncertainty. We investigate how people really do rank distributions in terms of welfare. According to Harsanyi, the evaluation can be done from the standpoint of an uninvolved external judge, a public official, for example, or by a person who knows that she holds one of the positions in society, with an equal chance for any of the available positions. Are these two structures to be viewed differently? We use a questionnaire experiment to focus on the two different interpretations of the Harsanyi approach. There are important, systematic differences that transcend the cultural background of respondents.     

Chartering strategies for shipping companies

This paper describes a model which may be used in the evaluation of the relative effectiveness of policies used in the areas of ship chartering and the switching of combined carriers between the dry cargo market and the tanker market. The policies for chartering and market switching are expressed in the form of ‘desired proportions’ of the fleet operating in a particular charter-mode or a particular market. Graphical illustration of this form of expression of policies can be easily understood, so that alternative policies can be designed with relative ease. The effectiveness of a particular form of policy under various freight market conditions can be determined from the model, the use of which is illustrated by its application to the deployment of the fleet of an hypothetical shipping company. It is assumed that the feedback connection between the company's actions and the marketplace is negligible.     

The effect of alternative staff time data collection methods on drug treatment service cost estimates

Although a limited number of service cost estimates exist, no study has evaluated how differences in the method used to collect the staff time allocation across treatment services contribute to differences in service cost estimates. Three alternative data collection methods for estimating service-level costs in methadone treatment programs were evaluated: key informants, staff surveys, and staff diaries. We analyzed data from 25 methadone clinics across the United States. Results indicate that for the three primary services offered at methadone clinics—individual counseling, group counseling, and methadone dosing—no statistically significant differences exist in the mean estimates of costs per session across programs. Of the other five services analyzed, we found no statistically significant differences in two of the mean costs per session and a small but statistically significant difference in another service. We found large and statistically significant differences in mean costs for two services, initial patient assessment and initial medical services. Although there is no gold standard available to judge which method is the best to use, we concluded that the key informant method yields more reliable cost estimates compared with the staff methods and is less burdensome to both the treatment programs and to researchers. Our findings suggest that the key informant method is the preferred method for costing substance abuse treatment services.     

Inequality decompositions—a reconciliation

We show how classic source-decomposition and subgroup-decomposition methods can be reconciled with regression methodology used in the recent literature. We also highlight some pitfalls that arise from uncritical use of the regression approach. The LIS database is used to compare the approaches using an analysis of the changing contributions to inequality in the United States and Finland.     

Distribution-Free Inference for Welfare Indices under Complete and Incomplete Information

The data available for estimating welfare indicators are often inconveniently incomplete data: they may be censored or truncated. Furthermore, for robustness reasons, researchers sometimes use trimmed samples. By using the statistical tool known as the Influence Function we derive distribution-free asymptotic variances for wide classes of welfare indicators not only in the complete data case, but also in the important cases where the data have been trimmed, censored or truncated.     

Reference distributions and inequality measurement

We investigate a general problem of comparing pairs of distributions which includes approaches to inequality measurement, the evaluation of “unfair” income inequality, evaluation of inequality relative to norm incomes, and goodness of fit. We show how to represent the generic problem simply using (1) a class of divergence measures derived from a parsimonious set of axioms and (2) alternative types of “reference distributions.” The problems of appropriate statistical implementation are discussed and empirical illustrations of the technique are provided using a variety of reference distributions.