Department of Statistics and Operations Research

Seminar Schedule: Year 2011


Rafael Romero Villafranca, Departamento de Estadística e Investigación Operativa Aplicadas y Calidad, Universidad Politécnica de Valencia.
  • Monday, June 6th, 2011. Time: 12:30
BALANCEDNESS OF HOMOGENEOUS OF DEGREE ONE GAMES WITH APPLICATIONS TO COOPERATIVE QUEUEING GAMES WITH OUTSOURCING
Shoshana Anily, Leon Recanati Graduate School of Business Administration of Tel Aviv University, Israel
Practical statistics consulting from ISRU: Newcastle University’s statistics outreach unit
SPEAKER: Shirley Coleman

LANGUAGE: English.

PLACE: Edifici C5, Aula: C5016, Campus Nord, UPC (see map)

DATE: Divendres, 16 de Desembre de 2011, Hora: 12:30

ABSTRACT: British Universities like to make an important contribution to business and society. ISRU has been supported by Newcastle University and valued for over 25 years. Working as a consultant within ISRU is extremely interesting and varied. We have obtained support funding to help local companies repair the damage of years of statistics-avoidance and to make use of statistics to improve their profitability. Much of ISRU’s work is devoted to training and we realise the value of good teaching skills. As companies try to save money, they often downsize staff numbers. ISRU provides statistical expertise where companies do not have suitably qualified staff. We have encountered many situations where statistical theory needs to be carefully adapted to provide practical conclusions and guidelines to help companies make important decisions. The talk will describe ISRU’s role and work and give illustrative examples of recent assignments.
The speaker:
click here for more information about Shirley Coleman.

Back to start


Vehicle Routing and Emissions/Risk minimization

SPEAKER:Irina Gribkovskaia/Oyvind Halskau. Molde University College, Norway

LANGUAGE: English.

PLACE: C6 Building, Room: C6003, Campus Nord, UPC (see map)

DATE: Thursday, December 15th, 2011. Time: 12:30

ABSTRACT:

In the offshore petroleum industry, employees are transported to and from the offshore installations by helicopter, which represents a major risk. In this talk we analyze how to improve transportation safety by solving the helicopter routing problem with a risk objective expressed in terms of expected number of fatalities. Three routing policies are considered: a direct routing policy, a Hamiltonian routing policy, and a general routing policy. We present and comment the results of a computational study conducted in order to assess and compare these policies under a travel time, a passenger risk and a combined passenger and pilot risk objective. In particular, our results show that passenger transportation risk can be reduced by increasing travel time at the expense of pilot risk. This can be achieved through a reduction of the average number of passengers on board by applying either a Hamiltonian or a general routing policy. Our methodology can also be used to derive an equitable distribution of risk between passengers and pilots, considering that pilots fly much more frequently than passengers.

***************

The travelling salesman problem is a well known combinatorial problem. Today we know enough to solve big instances to optimality with the help of different techniques. But still the problem belongs to the NP-complete problems. In this small talk I will present two statistical moments for the said problem, that under certain conditions can give some insights into the probability distribution for the TSP. Chebyshevs inequality will be applied in this context.

The speakers:

Irina Gribkovskaia is a Professor in Quantitative Logisitics from Molder University College, Specialized University in Logistics. Her background is in theory of optimal control and mathematical programming. She has a PhD degree in differential equations. During the last decade she has been conducting research in quantitative logistics with particular focus on operations research and combinatorial optimization. Her field of research is formulation of mathematical programming models and development of algorithms that can be used for the solution of many practical optimization problems of industrial logistics, in such areas as distribution, vehicle routing and fleet sizing, scheduling and production planning.

Oyvind Halskau is an assistant professor at Molde university college from 1988 and associate professor from 2000. He has a major in mathematics from the University of Oslo and a PhD in economy from the Norwegian School of business and Administration in Bergen. His main research has been within Vehicle Routing and inventory management.

click here for more information about the reseach group of the speakers

Back to start

Data Shuffling
SPEAKER:Krish Muralidhar

LANGUAGE: English.

PLACE: Edifici C5, Aula: C5016, Campus Nord, UPC (see map)

DATE: Friday, November 18th, 2011. Time: 12:30

ABSTRACT: This study discusses a new procedure for masking confidential numerical data-a procedure called data shuffling-in which the values of the confidential variables are "shuffled" among observations. The shuffled data provides a high level of data utility and minimizes the risk of disclosure. From a practical perspective, data shuffling overcomes reservations about using perturbed or modified confidential data because it retains all the desirable properties of perturbation methods and performs better than other masking techniques in both data utility and disclosure risk. In addition, data shuffling can be implemented using only rank-order data, and thus provides a nonparametric method for masking. We illustrate the applicability of data shuffling for small and large data sets.
The speaker:
click here for more information about Krish Muralidhar

Back to start


Quadratic Minimisation Problems in Statistics — Applications

SPEAKER: John Gower

LANGUAGE: English.

PLACE: FME, Edifici U, Aula: Sala d'Actes, Campus Sud, UPC, Pau Gargallo, 5, 02028 Barcelona (see map)

DATE: Divendres, 7 d'Octubre de 2011, Hora: 12:30

ABSTRACT: The problem minx(x-t)'A(x-t) subject to x'Bx+2b'x=k where A is positive definite or positive semi-definite commonly occurs in statistics. This problem will be discussed within the framework of a general unifying methodology which will be outlined. The generalization may be considered to give an extension to familiar eigenvalue formulations (t = 0, b = 0, A is pd and B is psd). Variants include non-trivial considerations that arise when (i) A and/or B are not of full rank and (ii) t takes special forms (especially t = 0 and (iii) when B is not definite. Geometrical interpretations give insight and will be provided. Applications discussed will include Canonical analysis (possibly in reduced rank form), optimal scaling, Oblique Procrustes analysis, Missing values in Procrustes analysis, quadratic forms, splines, Constrained regression, Hardy-Weinberg estimation, ALSCAL.
The speaker:
click here for more information about John Gower.

Back to start



An statistician at NICE

SPEAKER: Mike Capmbell, National Institute for Health and Clinical Excellence (NICE)

LANGUAGE: English.

PLACE: Edifici C5, Aula: per determinar, Campus Nord, UPC (see map)

DATE: Dilluns, 12 de setembre de 2011, Hora: 12:30

ABSTRACT:
The National Institute for Health and Clinical Excellence (NICE) is the UK's body that decides whether medical treatments should be paid
for by the National Health Service. Part of its job is to appraise new technologies for effectiveness and cost-effectiveness. The speaker has
sat on a NICE appraisal committee for 10 years as a statistician and will talk about how NICE comes to its decisions. Some of the interesting statistical issues that have come up over the years will
be highlighted, in particular sub-group analysis and the relevance of statistical significance.

The speaker:
click here for more information about Mike Campbell.

Back to start


THE STABLE SET POLYTOPE OF CLAW - FREE GRAPHS WITH STABILITY NUMBER GREATER THAN 3

SPEAKER: Claudio Gentile, Instituto di Analisi dei Sistemi ed Informatica "Antonio Ruberti" Consiglio Nazionale delle Ricerche, Itàlia

LANGUAGE: English.

PLACE: Edifici C5, Aula: C5016, Campus Nord, UPC (see map)

DATE: Dimecres, 22 de Juny de 2011, Hora: 12:30

ABSTRACT: In 1965 Edmonds provided the first complete description of the polyhedron associated with a combinatorial optimization problem: the matching polytope. As the matching problem is equivalent to the stable set problem over line graphs, many researchers tried to generalize Edmonds' result by considering the stable set problem over a super class of line graphs: the claw-free graphs. However, as testified also by Grötschel, Lovasz, and Schrijver in 1988 ``in spite of considerable efforts, no decent system of inequalities describing STAB(G) for claw-free graphs is known''.

In this work we provide an explicit linear description of the stable set polytope of claw-free graphs with stability number at least four and with no 1-join.

El ponent:

clicar aquí per accedir a la página personal de Claudio Gentile.

Back to start



ALGUNAS EXPERIENCIAS EN LA COLABORACIÓN DE UN ESTADÍSTICO CON LAS INDUSTRIAS DE ACERO
PARA LA CONSTRUCCIÓN
CONVIDAT: Rafael Romero Villafranca, Departamento de Estadística e Investigación Operativa Aplicadas y Calidad, Universidad Politécnica de Valencia.
IDIOMA: Castellà.
LLOC: Edifici C5, Aula: C5016, Campus Nord, UPC (see map)
DATA: Dijous, 9 de Juny de 2011, Hora: 12:30
RESUM: Desde hace 7 años vengo colaborando estrechamente con las industrias españolas y portuguesas de acero para la construcción, un sector cuya facturación anual es del orden de 5.000 millones de euros. En la charla comentaré algunos aspectos de esa colaboración, resaltando dos de ellos que han exigido el desarrollo de planteamientos originales e, incluso, alguna aportación novedosa de carácter metodológico. Por una parte me centraré en el tema de la estimación de los valores característicos a largo plazo para las propiedades mecánicas de los aceros para armaduras, una cuestión para la que las Normas aceptadas internacionalmente proponen una solución errónea y para la que tuve que desarrollar un procedimiento alternativo estadísticamente correcto. En segundo lugar comentaré el tema de la estimación de dichos valores característicos a corto plazo a nivel de coladas, una cuestión de cierta relevancia y para la que, en el desarrollo del Reglamento de la marca N de Aenor para acero corrugado, hemos desarrollado una solución original basada en un enfoque bayesiano del problema. Espero que podamos dedicar algún tiempo a intercambiar opiniones sobre el papel del consultor estadístico en las empresas y sobre el perfil humano y la formación adecuada para desarrollar esa labor.
El ponent: clicar aquí per accedir a la página personal de Rafael Romero.
Back to start


BALANCEDNESS OF HOMOGENEOUS OF DEGREE ONE GAMES WITH APPLICATIONS TO COOPERATIVE QUEUEING GAMES WITH OUTSOURCING
CONVIDADA: Shoshana Anily, Leon Recanati Graduate School of Business Administration of Tel Aviv University, Israel
IDIOMA: English.
LLOC: Edifici C5, Aula: C5016, Campus Nord, UPC (see map)
DATA: Dilluns, 6 de Juny de 2011, Hora: 12:30
RESUM: We consider a cooperative game with transferable utility, defined on a set of players N, whose characteristic function V depends only on a vector of some quantitative properties of the players, and is otherwise independent of their identity. As a result, the game can be defined on any virtual coalition of players, not necessarily subsets of N, as long as each player is associated with a vector of properties. We call such games regular games. Many games considered in the literature are regular. We assume that the characteristic function of the game is sub-additive, i.e., the cost of two disjoint coalitions is at least as large as the cost of their union, implying economies of scope. Sub-additivity implies that a natural outcome of any bargaining process is the formation of a single coalition N called the grand coalition. A central question in cooperative games is how to fairly allocate the cost V(N) among the players of N. The literature refers to a few concepts of fairness. Here we focus on the known concept of the core: core allocations guarantee that no coalition has an incentive to quit the grand coalition. A fundamental problem is that the core may be empty. The literature describes just a couple of methods for proving the nonemptiness of the core. In this talk we present what we believe is a novel condition that guarantees the non-emptiness of the core. We refer to a property called homogeneity of degree one, which means that when k coalitions, associated with exactly the same set of vectors of properties, cooperate, the resulting cost is k times the cost of a single coalition. Homogeneity of degree one implies lack of economies of scale. We prove that if the characteristic function of a regular game is sub-additive and homogeneous of degree one, then the core of the game is non-empty. We then present a few examples for such games which naturally emerge when servers in queuing systems cooperate in order to outsource some of the activities, while the rest of the activities are optimally assigned to the servers in the system. The existing conditions for showing the non-emptiness of the core fail here, while the conditions stated above are satisfied, proving that the cores of these games are non-empty.
This is a joint work with prof. Moshe Haviv from the Hebrew University of Jerusalem
La ponent:
Shoshana Anily holds her Ph.D. in Management Science from the Columbia Business School, an MA degree is statistics and B.Sc. degree in mathematics, both from Tel Aviv University. Upon graduation she joined the University of British Columbia, and in 1989 she returned to her home country and joined Tel Aviv University. She is a professor of decisions and Operations Research at the Leon Recanati Graduate School of Business Administration of Tel Aviv University, Israel. Her research interests include performance analysis of heuristics, routing problems, inventory control, production, scheduling, and cost allocation problems in a supply chains. She spent some periods as a visiting professor at Columbia University, EPFL and the CRT at the University of Montreal
Clicar aquí per accedir a la página personal de Shoshana Anily.
Inici
An Iterated Local Search for a Multiobjective Vehicle Routing Problem
a la página personal de Rafael Romero.

Back to start

An Iterated Local Search for a Multiobjective Vehicle Routing Problem

 

CONVIDAT: Helena Ramalhinho Lourenço.

IDIOMA: .

LLOC: Edifici C5, Aula: C5016, Campus Nord, UPC (see map)

DATA: Divendres, 27 de Maig de 2011, Hora: 12:30

RESUM: Problems arising in distribution of retailing industries are complex and involve several departments and decision levels. One of the most important decisions is the design of the routes to distribute the products to the stores in an efficient and inexpensive way but also satisfying marketing and logistics objectives. This leads to a multiobjective problem since the marketing and logistics objective when designing the routes can be conflicting. This chapter explores three different distribution routing strategies and proposes an Iterated Local Search (ILS) to solve the Multiobjective Vehicle Routing Problem. The first strategy corresponds to the classical vehicle routing problem where total distance or cost is minimized. The second strategy is a master route strategy with daily adaptations where customer loyalty is maximized, which is one of the objectives of the Marketing department. The third strategy takes into account the cross-functional planning between Logistics and Marketing department through a multi-objective model. The ILS is based on a local search for the TSP problem and extended to the VRP with more than one objective. An analysis and comparison of the three strategies is presented through a computational experiment. The cross-functional planning strategy leads to solutions that put in practice the coordination between the two functional areas of Marketing and Logistics and better meet business objectives in general.

This is a joint work with Antonio Ladrón de Guebarra and Rita Ribeiro.

El ponent:

clicar aquí per accedir a la web d'Helena Ramalhinho


Back to start

APLICACIONES DE LA ESTADÍSTICA A MATERIALES INGENIERILES

CONVIDAT: Salvador Naya Fernández, Grupo de Investigación MODES, Departamento de Matemáticas, Universidad de A Coruña.

IDIOMA: Castellà.

LLOC: Edifici C5, Aula: C5016, Campus Nord, UPC (see map)

DATA: Divendres, 13 de Maig de 2011, Hora: 12:30

RESUM: La estadística moderna se está enfrentando a problemas con estructuras de datos complejas que demandan el uso de métodos sofisticados y frecuentemente computacionalmente intensivos. En este contexto, el objeto de esta charla es mostrar algunos ejemplos de colaboración interdisciplinar, presentando una selección de aplicaciones recientes de nuevos métodos estadísticos a problemas surgidos en el ámbito de la ciencia de los materiales. Entre ellos, se expondrá cómo modelizar, mediantes técnicas de regresión tanto paramétricas como no paramétricas, distintas situaciones asociadas al proceso de degradación de materiales mediante pruebas de laboratorio y sus aplicaciones en distintos problemas ingenieriles. Concretamente se analizará un modelo de mezcla de funciones logísticas como alternativa al clásico modelo de Arrhenius para la modelización de la pérdida de masa de un material al variar su temperatura (curvas TGA) y su aplicación al recubrimiento de los depósitos de combustible del transbordador Challenger [1,2]. En esta misma línea de modelización de curvas termogravimétricas, se propondrá un nuevo contraste de hipótesis que permite verificar el ajuste de estos nuevos modelos [3].

Además de presentar el desarrollo de nuevos métodos estadísticos a partir de los resultados metodológicos recientes, también se analizará la aplicación de otros métodos recientemente propuestos, a diversos problemas relevantes en este campo de la ciencia de los materiales. Concretamente se expondrán las ventajas que aporta el tratamiento del estudio de los datos funcionales para estimar propiedades físicas de materiales compuestos. Mediante esta herramienta estadística se presentará la aplicación de un diseño de experimentos usando el modelo del ANOVA funcional, para estudiar la influencia que tiene la adición de una mezcla de nano y micro partículas de humo de sílice, subproducto de la industria del silicio, en la degradación térmica de una resina epoxi [4]. Finalmente, se presentará una aplicación dentro del campo de reconocimiento de imágenes, la intención es ver cómo mediante técnicas estadísticas de clasificación, como el clásico análisis discriminante o técnicas más recientes como las SVM (Support Vector Machines) es posible diferenciar distintas especies de madera en base a imágenes obtenidas por microscopía SEM (Scanning Electron Microscopy) [5].

Referencias

1. Naya, S., Cao, R., López-de-Ullibarri, I., Artiaga, R., Barbadillo, F. y García, A. Logistic mixture model vs Arrhenius for kinetic study of degradation of materials by dynamic thermogravimetric analysis. Journal of Chemometrics, 20, 158-163. 2006.

2. Naya, S, Martínez-Vilariño, S. y Artiaga. R. Effects of thermal cycling on permeability and thermal properties of nanoclay-epoxy composites. DYNA (www.revistadyna.com). 84, 2. 151-156. 2009.

3. Cao, R. y Naya, S. Nonlinear regression checking via local polynomial smoothing with applications to thermogravimetric analysis. Journal of Chemometrics. 23, 6, 275- 282. 2009.

4. Tarrío, J., Naya, S., Francisco-Fernández, M., Artiaga, R y López-Beceiro, J. Application of functional ANOVA to the study of thermal stability of micro-nano silica epoxy composites. Chemometrics and Intelligent Laboratory Systems 105, 114-124. 2011,

5. Tarrío, J., Naya, S., Francisco-Fernández, M., Artiaga, R y López-Beceiro, J. Functional nonparametric classification of wood species from thermal data. Journal of Thermal Analysis and Calorimetry. Publicado online 8 diciembre 2010. DOI10.1007/s10973-010-1157-2.

Back to start

Han mort les metaheurístiques?: Visquin les matheurístiques!

CONVIDAT: Albert Corominas, Institut d'Organització i Control de Sistemes Industrials, Departament d'Organització d'Empreses, UPC

IDIOMA: Catalan

LLOC: Edifici C5, Aula: C5016, Campus Nord, UPC (see map)

DATA: Divendres, 8 d'Abril de 2011, Hora: 12:30

RESUM: En aquesta sessió del seminari es tractarà de donar respostes a les preguntes següents:

  • Amb quins criteris hem de decidir com abordar la resolució d'un problema d'optimització combinatòria?
  • Si un problema és "dur", hem de renunciar d'entrada a resoldre'l òptimament?
  • Està justificat l'ús quasi sistemàtic de metaheurístiques sofisticades?
  • Si els procediments exactes fallen, hem d'abandonar les matemàtiques o és millor recórrer a les matheurístiques?
  • Què són i quines són les més significatives?

El ponent: clicar aquí per accedir a la web d'Albert Corominas

Back to start

Solving the Single vehicle routing problem with variable capacity

CONVIDAT: François Louveaux, Département des Méthodes Quantitatives, Facultés Universitaires Notre Dame de la Paix, Namur.

IDIOMA: English.

LLOC: Edifici C5, Aula: C5016, Campus Nord, UPC (see map)

DATA: Divendres, 1 d'Abril de 2011, Hora: 12:30

RESUM: In the capacitated vehicle routing problem (VRP), the objective is to find a set of delivery routes, starting and ending at the depot, of minimal total length. Customers have a known demand and the total demand on each route may not exceed a given capacity Q. As a consequence, the vehicle may have to return a few times to the depot in order to unload and resume its trip. Alternatively, several vehicles are needed. In the stochastic version (SVRP), some customer’s demands are stochastic with a known distribution; the actual value is revealed only when the vehicle arrives at the customer. When some customer is visited, the vehicle may be unable to deliver the demand. Such a situation is called a failure and requires some recourse action (e.g., returning to the depot to reload the vehicle). All these versions assume a vehicle with given size. In the VRP, selecting a vehicle with a larger size may reduce the number of returns to the depot (or the number of vehicles). Similarly, a larger vehicle size in the stochastic SVRP may reduce the risk of failure. It is clear however that larger vehicles are more costly, both to acquire as to operate. Thinking in terms of cost minimization (and thus not only in terms of distance minimization), a trade-off is clearly desirable there. In this paper, we propose formulations for this problem, linearization of the capacity constraints and solution methods. We also provide some computational results

This is a joint work with Juan José Salazar-González

El ponent:

clicar aquí per accedir a la web de François Louveaux

Back to start

ESTIMATION OF SURVIVAL CURVES IN THE CASE OF "TIME VARYING TREATMENT"

CONVIDAT: Urania Dafni, Department of Nursing, Laboratory of Biostatistics, University of Athens.

IDIOMA: English.

LLOC: Edifici C5, Aula: C5016, Campus Nord, UPC (see map)

DATA: Divendres, 11 de Març de 2011, Hora: 12:30

RESUM: Estimating survival curves in a non-parametric way via the use of Kaplan-Meyer probabilities is common practice. In the clinical trial setting, in the case of randomized treatments, this approach is widely applied and its properties are well understood. In cases where the treatment varies across time (e.g. switching to another therapy) attempts have been made based on the classical Kaplan-Meyer to estimate the survival curve for the group of patients that changed treatment.

For these cases the time varying Cox model can estimate efficiently and without bias the treatment effect, but no information about the survival curves is given. The aim of this work is to fill the gap by considering an appropriate estimate for the survival curve in this case.

An alternative estimate for the survival curve based on similar conditional arguments as in the typical Kaplan-Meyer but taking into account the survival and treatment history of the patients, is developed. Properties of the proposed estimator are discussed. Simulation evidence is also provided. The proposed survival estimates allow the extraction of information on the treatment effect for those that changed treatment and hence provide an improved estimate for the differential treatment benefit. The usage of these curves is exploited for deriving more accurate information on the effects of two initially randomized treatments, making use of the whole history of the patients.

The relationship of the new estimator with the Cox time varying model approach will be discussed. The proposed estimator can supplement the time varying Cox model as it allows its proper visualization.

Real data examples will be discussed as well as the potential of the new approach for improving the design of trials when switching treatment is possible. In this case the existing design ignoring this aspect leads to reduced power at the end of the study.

References

1. Simon, R. and Makuch, R.W. (1984) A non-parametric graphical interpretation of the relationship between survival and the occurrence of an event: application to responder and non-responder bias. Statistics in Medicine, 3, 35-44

2. Snapinn, S.M, Jiang, Q. and Iglewitz, B. (2005) Illustrating the impact of time varying covariate with an extended Kaplan-Meier estimator. The American Statistician, 59, 301-307

Back to start

STATISTICAL ASPECTS IN FORENSIC AND PATERNITY TESTING

CONVIDAT: Francesc Calafell, Institut de Biologia Evolutiva (CSIC-UPF), Universitat Pompeu Fabra, Barcelona.

IDIOMA: English.

LLOC: Edifici C5, Aula: C5016, Campus Nord, UPC (see map)

DATA: Divendres, 4 de Març de 2011, Hora: 12:30

RESUM: We often hear in the news that a DNA test has been performed to determine a paternity or to solve a crime. In this seminar, we will discuss what are the biological bases for forensic genetic tests, and, more importantly, we will deal with what CSI (Crime Scene Investigation) most often gets wrong: the statistical analysis behind the tests. We will present how the evidence is given, usually in a likelihood ratio framework, and we will see how the LR is computed and applied in different situations.

Back to start

Optimal Contract Pricing of Distributed Generation in Distribution Networks.

CONVIDAT: Javier Contreras, Departamento de Ingeniería Eléctrica, Universidad de Castilla- La Mancha..

IDIOMA: Spanish

LLOC: Aula: C4002-Aula amfiteatre, Campus Nord, UPC (see map) <--- Compte : lloc no habitual

DATA: Divendres, 14 de gener de 2011, Hora: 12:30

RESUM: This paper proposes a bilevel programming approach to determine the optimal contract price of dispatchable distributed generation (DG) units in distribution systems. Two different agents are considered in the model, namely, the distribution company (DisCo) and the owner of the DG. The former seeks the minimization of the payments incurred in attending the forecasted demand, while the latter seeks the maximization of his profit. To meet the expected demand, the DisCo has the option to purchase energy from any DG unit within its network and directly from the wholesale electricity market. A traditional distribution utility model with no competition among DG units is considered. The proposed model positions the DG owner in the outer optimization level and the DisCo in the inner one. This last optimization problem is substituted by its Karush-Kuhn-Tucker optimality conditions, turning the bilevel programming problem into an equivalent single-level nonlinear programming problem which is solved using commercially available software. Tests are performed in a modified IEEE 34-bus distribution network.

El ponent:

clicar aquí per accedir a la web de Javier Contreras.

Back to start