syllabi M2 Stat_Eco - L`École d`économie de Toulouse
Transcription
syllabi M2 Stat_Eco - L`École d`économie de Toulouse
Master 2 Economics & Statistics 2015-2016 SYLLABUS Big Data Survey Sampling Time Series Lifetime Analysis Econometric Panels Econometrics of Qualitative Variables Marketing Non Parametric Models Linear Models Statistical Consulting Workshop Statistical Software Life Insurance Non Life Insurance Spatial Econometrics Geomarketing Data Bases Web Mining Data Mining Scoring 2015-2016 Intitulé du cours Niveau / Semestre Crédits Enseignant responsable Autre(s) enseignant(s) Volume horaire CM Volume horaire TD Volume horaire TP Big Data Master 2 Economics & Statistics/semester 1 6 S. GADAT 36 0 0 OBJECTIVES This course provides a modern view of statistical methods for high dimensional estimation and classification, as well as stochastic prediction. For each method, we will describe the underlying theoretical aspect with associated statistical and numerical results. The course will take place in a computer room so that we can illustrate immediately the ideas covered in lecture through simulated and real data sets. The course notes, data sets, exercises and their solutions, R and Matlab codes, and other materials will be made available for the students. COURSE OUTLINE Reminders on Convex Optimization (1 session) High Dimensional regression (2 sessions) Stochastic Algorithms (1 session) High dimensional Classification (1 session) Matrix completion (1 session) REQUIREMENTS R & Matlab programming knowledge TEXT BOOKS Bühlmann, P. and van de Geer, S. (2011). Statistics for High-Dimensional Data: Methods, Theory and Applications. Springer. GRADING POLICY UE 1 2015-2016 Intitulé du cours Niveau / Semestre Crédits Enseignant responsable Autre(s) enseignant(s) Volume horaire CM Volume horaire TD Volume horaire TP Survey Sampling Master 2 Economics & Statistics/semester 1 3 A.RUIZ-GAZEN 18 0 0 OBJECTIVES Sampling theory course gives the basis for inference in a design based approach. Different sampling designs and estimation methods are studied in detail. COURSE OUTLINE Chapter 1: Introduction Chapter 2: Estimation of totals and means using the Horvitz-Thompson estimator for a simple random sampling with and without replacement. Calculation of a sample size for a given precision. Chapter 3. Estimation of totals and means using the Horvitz-Thompson estimator for a general sampling design. Bernoulli and unequal probabilities with replacement survey designs. Design effect. Chapter 4: Estimation of a ratio. Estimation by substitution of a ratio and estimation by linearization of its asymptotic variance. Chapter 5: Stratified sampling. Allocation of the total sampling size to the strata. Horvitz-Thompson estimation and variance estimation. Chapter 6: Post-stratified estimation, estimation by ratio and estimation by regression. Estimation methods which take into account auxiliary information. Link with calibration estimators. REQUIREMENTS TEXT BOOKS GRADING POLICY The course is assessed through an interrogation that takes place mid-course, a final exam and a project to hand in. UE 2a 2015-2016 Intitulé du cours Niveau / Semestre Crédits Enseignant responsable Autre(s) enseignant(s) Volume horaire CM Volume horaire TD Volume horaire TP Time Series Master 2 Economics & Statistics/semester 1 3 F.BARTHE 18 0 0 OBJECTIVES The objective of this module is twofold : provide the students with a good understanding of the main probabilistic models used in the statistic analysis of time series, and apply these models to concrete data. COURSE OUTLINE Times series analysis deals with the description, modeling and forecasting of data which are collected along time and enjoy strong correlations between successive observations. After reviewing the classical methods to extract the trend and seasonality of a time series, we will focus on the study of the residual/noise term. We will introduce the main notions in the study of weakly stationary random processes (autocorrelation, partial autocorrelation, spectral density, innovation). The AutoRegressive Moving Average processes will be presented, together with the methods allowing to identify them from observations. Eventually, models with autoregressive conditional heteroskedasticity will be considered, as they are employed commonly in modeling financial time series that exhibit time-varying volatility clustering. The class also includes sessions on machines to implement all of these methods through the use of the R software REQUIREMENTS Probability and Statictics course of bachelor level. Basic knowledge on complex numbers. TEXT BOOKS • BROCKWELL P. & DAVIS R. « Introduction to time series and forecasting » Springer • GOURIEROUX C. et MONFORT A. – « Cours de séries temporelles », Economica • ARAGON Y. – « Séries temporelles avec R», Springer More on: http://www.math.univ-toulouse.fr/~barthe/M2timeseries GRADING POLICY Final exam and short report to hand after the final session on machines. UE 2b 2015-2016 Intitulé du cours Niveau / Semestre Crédits Enseignant responsable Autre(s) enseignant(s) Volume horaire CM Volume horaire TD Volume horaire TP Lifetime Analysis Master 2 Economics & Statistics/semester 1 2, 5 E. LECONTE 18 0 0 OBJECTIVES To recognize a situation where censored data appear and to be able to analyze such data by implementing an appropriate modelling. All the courses take place in a computer room and an important place is given to the learning of the dedicated software (R and SAS). COURSE OUTLINE Survival distributions: specific functions and the most used distributions for survival data. Censored data: censoring and truncation, right censoring. Non parametric estimation of the survival function and the cumulative hazard function: Kaplan- Meier and Nelson- Aalen estimators. Comparison of two or more survival distributions : the weighted logrank tests. Parametric regression models: proportional hazards models and accelerated failure time models. The semi-parametric Cox model: modelling, partial likelihood, estimation and tests of the parameters, estimation of the baseline cumulative hazard, validation and extensions. REQUIREMENTS Inferential statistics: hypothesis testing, maximum likelihood estimation and tests (Wald, score and likelihood ratio tests). TEXT BOOKS Hill C., Com-Nougué C., Kramar C., Moreau T., O'Quigley J., Senoussi R. and Chastang C. (1990), Analyse statistique des données de survie, Flammarion. Klein J. P. and Moeschberger M. L. (1997), Survival Analysis - Techniques for censored and truncated data, Springer. Allison P. D. (1995), Survival analysis using the SAS system. A practical guide. SAS Institute Inc. GRADING POLICY A final exam with a short theoretical part without documents and a longer part to analyse real data with computer software (R or SAS) with all the documents allowed. UE 3a 2015-2016 Intitulé du cours Niveau / Semestre Crédits Enseignant responsable Autre(s) enseignant(s) Volume horaire CM Volume horaire TD Volume horaire TP Econometric Panels Master 2 Economics & Statistics/semester 1 2, 5 P. LAVERGNE OBJECTIVES Understand and be able to apply econometric methods for panel data. COURSE OUTLINE Background on linear regression Panel data Instrumental variables REQUIREMENTS Linear algebra, probability, statistical inference, basic econometrics. TEXT BOOKS Introduction to Econometrics by J. Stock and M. Watson Pearson Education Principes d'Econometrie : traduction française J. Trabelsi GRADING POLICY Project 40%, final exam 60% 18 0 0 UE 3b 2015-2016 Intitulé du cours Niveau / Semestre Crédits Enseignant responsable Autre(s) enseignant(s) Volume horaire CM Volume horaire TD Volume horaire TP Econometrics of Qualitative Variables Master 2 Economics & Statistics/semester 1 2, 5 L. BONNAL 18 0 0 OBJECTIVES Construct an explanatory model for a qualitative variable or a truncated variable. COURSE OUTLINE Chapter 0: Review of qualitative variables, indicator and maximum likelihood estimation Chapter 1: The simple qualitative model: logit and simple probit. Maximum likelihood estimation, marginal effects, odds ratio. Chapter 2: The bivariate qualitative model. Estimation, problem identification, calculations ATE, ATT, ATNT. Chapter 3: The multinominal qualitative model. Ordered model, unordered model, sequential model. Chapter 4: The truncated or censored models: single and generalized Tobit. REQUIREMENTS Econometrics of linear models, statistical testing theory. TEXT BOOKS Econometrics analysis of cross section panel data years, Wooldridge, MIT press Econometrics of qualitative variables, A Thomas, Dunod edition Limited dependent and qualitative variables in econometrics, Maddala, Econometric Society monographs. GRADING POLICY Final exam on machines. UE 3c 2015-2016 Intitulé du cours Niveau / Semestre Crédits Enseignant responsable Autre(s) enseignant(s) Volume horaire CM Volume horaire TD Volume horaire TP Marketing Master 2 Economics & Statistics/semester 1 2, 5 P. BIZZARI 18 0 0 OBJECTIVES Courses on the operational aspects of statistical techniques in the context of marketing orientated data mining projects. The purpose of this course is to raise awareness of the ins and outs of a project in a professional environment without being exclusively focused in either the technical or theoretical part. This is illustrated in two stages via a presentation of the main methodologies used (score, segmentation, decision tree.... etc). Then via a practical project in groups. COURSE OUTLINE Marketing relationship and CRM Definitions and concepts, structure of information systems (data warehouse), data mining issues Data mining mapping techniques Statistics and data mining, application areas of data mining Process of conducting a study Conceptualisation, implementation, integration, measurement and monitoring Data mining perspectives Trends and issues of tomorrow Case studies - Realisation of a project Target marketing, customer segmentation, product associations, scoring attrition ("churn"), calculating customer value REQUIREMENTS TEXT BOOKS GRADING POLICY Final exam UE 3d 2015-2016 Intitulé du cours Niveau / Semestre Crédits Enseignant responsable Autre(s) enseignant(s) Volume horaire CM Volume horaire TD Volume horaire TP Non Parametric Models Master 2 Economics & Statistics/semester 1 2, 5 A.DAOUIA 18 0 0 OBJECTIVES This course provides a modern view of the most popular nonparametric methods, especially on the important topics of density and regression estimation in both univariate and multivariate cases. The basic idea of those methods is to let the data speak for themselves without recourse to any a priori parametric specification. For each method, we will introduce the underlying theoretical aspects (relevant mathematical results) although the proofs will be skipped. We will spend more time on cultural aspects (knowledge of the methodology and interpretation of statistical results), computational aspects (implementation using R and Matlab softwares), and case studies (returns of education, assets returns, etc). The course will take place in a computer room so that we can illustrate immediately the ideas covered in lecture through simulated and real data sets. The course notes, data sets, exercises and their solutions, R and Matlab codes, and other materials will be made available on the university's course website. COURSE OUTLINE Parametric versus nonparametric models Nonparametric density estimation Review of polynomial spline functions Nonparametric regression estimation Semi-parametric regression estimation REQUIREMENTS Traditional asymptotic statistics, Basics of R and Matlab programming. TEXT BOOKS Wasserman, L. (2006): All of Nonparametric Statistics. Springer GRADING POLICY Non parametric models: there will be an exam (50%) and a project (50%) in this class UE 3e 2015-2016 Intitulé du cours Niveau / Semestre Crédits Enseignant responsable Autre(s) enseignant(s) Volume horaire CM Volume horaire TD Volume horaire TP Statistical Consulting Workshop Master 2 Economics & Statistics/semester 1 5 36 0 0 OBJECTIVES The statistical consulting class is not a classical course but rather a team project. Teams of four students and two teachers are assigned a project topic given ny an outside pattern (firms, administrations, researchers in other fields) and work on the project in a collaborative way from the beginning of October to the end of March. COURSE OUTLINE The team meets once a week to discuss the advancement of the project and presents regularily the results to the outside member. REQUIREMENTS TEXT BOOKS GRADING POLICY An oral defense is organised at the end. UE 3g 2015-2016 Intitulé du cours Niveau / Semestre Crédits Enseignant responsable Autre(s) enseignant(s) Volume horaire CM Volume horaire TD Volume horaire TP Non Life Insurance Master 2 Economics & Statistics/semester 2 2,5 R. AIT‐MANSOUR 18 0 0 OBJECTIVES Apporter les connaissances et les compétences de base en actuariat de l’assurance non-vie permettant aux étudiants de poursuivre une carrière professionnelle en qualité de chargé d’études actuarielles, en compagnie d’assurance ou en cabinet de conseil. COURSE OUTLINE 1– Généralités sur l’assurance non-vie Ce premier chapitre est consacré à une présentation rapide de l’assurance non vie et ses spécificités 2– La tarification Ce second chapitre traite de la problématique de fixation d’un tarif pour un produit d’assurance non-vie: sur quels critères se calcule un tarif? Comment établir la tarification? 3- Le provisionnement Dans ce chapitre, la question de la constitution de réserves est abordée, tant pour répondre à une exigence comptable que pour respecter les engagements pris envers les assurés. 4- La réassurance Les divers dispositifs de réassurance sont abordés dans ce chapiteez: réassurance proportionnelle et réassurance non proportionnelle. 5- Eléments de comptabilité des assurances Dans ce chapitre, nous abordons les principes d'établissement des comptes d'assurance non-vie (bilan et compte de résultat). 6- La solvabilité Les régles prudentielles en viguer (norme Solvency 1) sont décrits dans ce chapitre. Les grandes orientations des futures régles (Solvency II) sont abordées en fin de chapitre. Les différents chapitres comportent des apllications numériques permettant d'illustrer les divers concepts abordés dans le cours. REQUIREMENTS Cours de Statistiques et de Probabilités de Licence. TEXT BOOKS TOSETTI Alain et Al.– « Assurance, Comptabilité, Réglementation, Actuariat », Economica DENUIT Michel et CHARPENTIER Arthur– « Mathématiques de l’assurance non-vie »,Economica PETAUTON Pierre– « Théorie de l’assurance dommages », Dunod GRADING POLICY UE 7b 2015-2016 Intitulé du cours Niveau / Semestre Crédits Enseignant responsable Autre(s) enseignant(s) Volume horaire CM Volume horaire TD Volume horaire TP Spatial Econometrics Master 2 Economics & Statistics/semester 2 2,5 A.RUIZ-GAZEN T.LAURENT, N.DEBARSY 18 0 0 OBJECTIVES The aim of this course is to introduce the statistical techniques adapted to the treatment of georeferenced data sets. COURSE OUTLINE The spatial dimension must be taken into account for modeling the inhomogeneity effects of a spatial process as well as its spatial dependence structure. After describing the three types of spatial data (geostatistical, areal and point patterns), we focus on the case of areal data which is more frequent in spatial econometrics. We present exploratory techniques dedicated to spatial data sets. We introduce specific tools such as variograms, weight matrices, Moran index, etc. for studying the autocorrelation structure. Then we introduce the family of spatial simultaneous autoregressive models (Durbin, spatial lag model, spatial error model). In the framework of these models, we discuss estimation and testing problems, interpretation of coefficients and marginal effects, and prediction. If time allows, we briefly cover some additional topics such as kriging and point patterns characteristics. The illustration of these techniques is done with R and the following packages: spdep, GeoXp, geoR, spatstat. REQUIREMENTS Linear models and R programming TEXT BOOKS Applied Spatial Data Analysis with R Roger S. Bivand, Edzer Pebesma and V. Gómez-Rubio UseR! Series, Springer GRADING POLICY Project UE 7c 2015-2016 Intitulé du cours Niveau / Semestre Crédits Enseignant responsable Autre(s) enseignant(s) Volume horaire CM Volume horaire TD Volume horaire TP Geomarketing Master 2 Economics & Statistics/semester 2 2,5 A.RUIZ-GAZEN A.LARA 18 0 0 OBJECTIVES This course introduces the techniques and concepts used in geomarketing. Geomarketing is the integration of the spatial dimension in the statistical marketing analyses. COURSE OUTLINE After defining the various concepts of trading areas, potential, we introduce the problems of turnover prediction and best location problems. For the former, we discuss the various interaction models used for flow data (gravity model, Huff model, Multiplicative competitive interaction model) and their estimation methods. For the latter, we review the literature on locationallocation problems. An applied part of the course taught by a professional illustrates the treatment of a geomarketing case study. REQUIREMENTS R programming TEXT BOOKS GRADING POLICY Project UE 7d 2015-2016 Intitulé du cours Niveau / Semestre Crédits Enseignant responsable Autre(s) enseignant(s) Volume horaire CM Volume horaire TD Volume horaire TP Data Bases Master 2 Economics & Statistics/semester 2 2,5 R. TOURNIER 18 0 0 OBJECTIVES The objective of this course is to earn a basic knowledge on Decision Support Systems through database systems. This course allows understanding how some computer tools may help a decision maker/manager to have a global vision of what is happening within his institution or company. The course introduces software used in the domain and details the use databases as well as On-Line Analytical Processing tools (also called OLAP tools). The course is an introduction to more the complex environment of decision support systems. COURSE OUTLINE Introduction to decision support and decision support systems (as well as a word on pivot tables and Excel—the most used tool); Query languages for databases (SQL). Starting with simple queries to the more complex analytical queries (application with Microsoft Access); OLAP analytical tools, design of multidimensional databases and analytical reports (application with SAP Business Objects). REQUIREMENTS None. Knowledge of databases and spreadsheets (Excel or Calc) may help. TEXT BOOKS Ralph Kimball books such as «The Data Warehouse Toolkit: The Complete Guide to Dimensional Modeling», 2008. GRADING POLICY Project done in pairs (40% of the evaluation) & final written exam (60% of the evaluation). UE 7e 2015-2016 Intitulé du cours Niveau / Semestre Crédits Enseignant responsable Autre(s) enseignant(s) Volume horaire CM Volume horaire TD Volume horaire TP Web Mining Master 2 Economics & Statistics/semester 2 2,5 Y.PITARCH 18 0 0 OBJECTIVES The objective of the module « web mining » is two-fold: it will be firstly to draw the widest theoretical overview of “web mining” and techniques conventionally used to search such data. Secondly, students will be confronted practically with this knowledge extraction problem via the establishment of a TP/project. COURSE OUTLINE It is customary to divide this area into three sub-areas: web content search (web content mining) structures search (web structure mining) and web usage search (web usage mining). For each of these categories, we describe the motivations, the applications, some work of major research and conclude by mentioning some challenges. Concerning the second part of the module, the objective is to allow students to experience all the knowledge extraction process: from data collection to analysis and interpretation of analytical results. Practical work will therefore have a purpose and students will be asked to provide a summary at the end of the module. Some basics of programming will be required (mainly to retrieve data and pre-treat it). A little booklet will be made available to students to help them in this task. The programming language used is yet to be determined but will probably be Perl. The software used for data analysis is not yet defined. At least two solutions seem relevant: the use of R coupled with data mining module or use of composition workflow software for extraction, pre-processing, processing, post-processing, data visualisation and analysis, e.g. Knime or Orange. REQUIREMENTS TEXT BOOKS Ralph Kimball books such as «The Data Warehouse Toolkit: The Complete Guide to Dimensional Modeling», 2008. GRADING POLICY UE 7f 2015-2016 Intitulé du cours Niveau / Semestre Crédits Enseignant responsable Autre(s) enseignant(s) Volume horaire CM Volume horaire TD Volume horaire TP Data Mining Master 2 Economics & Statistics/semester 2 6 X. GENDRE 36 0 0 OBJECTIVES The objective of this module is to introduce statistical methods in order to explore data structures. The extraction of information is central to the new challenges introduced by access to ever larger databases and this set of methods finds its interest in studies where potentially large scale data intervenes. COURSE OUTLINE In the first part of the course, we introduce the classic optimisation methods that are PCA and its variants (CA, MCA, MDA,...). Our approach is intentionally general in order to illustrate the mathematical principles common to all these tools which could easily be incorporated into a variety of settings. The second part of the course is an overview of methods commonly used in the industry to explore and process data. We talk about ecision trees (CART for example), neutral network (perceptrons, Kohonen networks), moving averages and ascending hierarchical classification. This class also includes sessions on machines to implement all of these methods through the use of th eR software. In addition, during these practice sessions, we will have the opportunity to introduce some more advanced points such as model selection, cross-validation, the bootstrap... REQUIREMENTS TEXT BOOKS GRADING POLICY UE 8 2015-2016 Intitulé du cours Niveau / Semestre Crédits Enseignant responsable Autre(s) enseignant(s) Volume horaire CM Volume horaire TD Volume horaire TP Scoring Master 2 Economics & Statistics/semester 2 6 36 0 0 OBJECTIVES Learn the background of generalized linear models and the basic skills of supervised classification using logistic regression. COURSE OUTLINE Theoretical part. After defining the scoring goals and vocabulary, the first chapter of the course is an introduction to the generalized linear model. Then we cover extensively the logistic regression model. We detail the definition, estimation and interpretation of the odds ratios as well as the computations of marginal effects. We discuss tools such as the Lorenz and ROC curves for the evaluation of the quality of the model and the selection of a score threshold. Applied part. A professional discusses a scoring case study. REQUIREMENTS R and SAS programming TEXT BOOKS Extending the linear model with R, J.J. Faraway, Chapman& Hall/CRC, 2006. J.M. Hilbe, Logistic regression models, CRC Press, Chapman and Hall, 2009. R. Anderson, The credit scoring toolkit, Oxford U.P., 2007. GRADING POLICY Exam and project. UE 9