Random May 2024

random@lists.dm.unipi.it

52 participants
70 discussions

Seminari MAD-la MAtematica dei DAti - Pisa
by Marco Romito 27 May '24

27 May '24

Carə tuttə, Siamo lieti di annunciare il seguente ciclo di incontri presso il Dipartimento di Matematica dell'Università di Pisa, organizzato nell'ambito del progetto speciale per la didattica /MAD - La Matematica dei Dati!/ *Martedì 28 maggio 2024, dalle ore _10:00_* Presso: Aula Magna *Francesco Preta (freelance AI researcher and consultant)* /"Basics of Natural Language Processing and Applications of Large Language Models"/ La presentazione sarà seguita da un tutorial con esempi di implementazione. *Martedì 4 giugno 2024, dalle ore 9:00* Presso: Aula Seminari *Juliette Achddou (Università degli Studi di Milano)* /"Introduction to Online learning: the multi-armed bandit problem"/ *Luca Papariello (BIP – Business Integration Partners S.p.A.)* /"Data-Driven Decision Making: Case Studies and Insights"/ *Daniele Malpetti (IDSIA)* /"Centralized and federated machine learning in biomedicine"/ / / Qui di seguito il link al sito con tutte le informazioni aggiornate:https://sites.google.com/unipi.it/mad/home/prossimi-eventi <https://sites.google.com/unipi.it/mad/home/prossimi-eventi>. Tutte le persone interessate sono invitate a partecipare. A presto, Andrea Agazzi Mario Correddu Aikaterini Papagiannouli Marco Romito ------------------------------------------------------------------------ Dear all, We are pleased to announce the following series of talks, that wil take place at the Math Department of the University of Pisa, organized as part of the MAD project - /The Mathematics of Data!/ *Tuesday, May 28, 2024, at_ 10:00_ AM *Location: Aula Magna *Francesco Preta (freelance AI researcher and consultant)* /"Basics of Natural Language Processing and Applications of Large Language Models"/ The presentation will be followed by a tutorial with implementation examples. *Tuesday, June 4, 2024, from 9:00 AM* Location: Aula Seminari *Juliette Achddou (Università degli Studi di Milano)* /"Introduction to Online learning: the multi-armed bandit problem"/ *Luca Papariello (BIP – Business Integration Partners S.p.A.)* /"Data-Driven Decision Making: Case Studies and Insights"/ *Daniele Malpetti (IDSIA)* /"Centralized and federated machine learning in biomedicine"/ / / The site of the project with all the updated information: https://sites.google.com/unipi.it/mad/home/prossimi-eventi <https://sites.google.com/unipi.it/mad/home/prossimi-eventi>. Anyone interested is welcome to join. Best regards, Andrea Agazzi Mario Correddu Aikaterini Papagiannouli Marco Romito

1 0

COPA 2024 @ Polimi | Tutorial Speakers Announced
by Simone Vantini 27 May '24

27 May '24

Dear all, We are pleased to announce the speakers for the tutorials at this year’s Symposium on Conformal and Probabilistic Prediction with Applications (COPA 2024) that will be held in Milan, Italy, at the Politecnico di Milano from the 9th to the 11th of September 2024. COPA is the annual event dedicated to the theory, methodology, and application of conformal and probabilistic prediction. It brings together researchers and practitioners from around the globe. COPA 2024 promises to be a hub of cutting-edge research, insightful discussions, and networking opportunities. Tutorial #1: Conformal Prediction in Python with crepes What to Expect: Discover the capabilities of crepes, a Python package that transforms standard classifiers and regressors into well-calibrated tools with reliable p-values and prediction sets. Prof. Henrik Boström will guide you through the core algorithms and demonstrate practical applications of conformal classifiers, regressors, and predictive systems using crepes. Tutorial Speaker: Prof. Henrik Boström from KTH Royal Institute of Technology, Stockholm, Sweden, is a leading expert in machine learning algorithms. With a focus on conformal prediction, ensemble learning, and explainable machine learning, Henrik has significantly contributed to various industries, including pharmaceuticals, healthcare, automotive, and insurance. His editorial and conference roles further underscore his influence in the field. Tutorial #2: MAPIE - Uncertainty Quantification Made Easy What to Expect: Explore MAPIE, a versatile Python library that simplifies the implementation of conformal prediction methods. Dr. Thibault Cordier will show you how MAPIE, part of the scikit-learn-contrib project, can handle tasks from classification and regression to more complex applications like multi-label classification and semantic segmentation, ensuring probabilistic guarantees on key metrics. Tutorial Speaker: Dr. Thibault Cordier from Lab Invent of Capgemini Invent, France, is a Data and Research Scientist leading the MAPIE project. With a PhD in Computer Science, his research focuses on distribution-free inference and conformal prediction, applied across computer vision, natural language processing, and time series analysis. For more details and to register, please visit the COPA 2024 Website at https://copa-conference.com/. Early bird registration fees are available until May 30, 2024. We look forward to seeing you in Milan! Kind regards, Simone Vantini Associate Professor of Statistics MOX Laboratory for Modeling and Scientific Computing Department of Mathematics, Politecnico di Milano simone.vantini(a)polimi.it<mailto:simone.vantini@polimi.it> Matteo Fontana Lecturer in Data Science Department of Computer Science School of Engineering, Physical and Mathematical Sciences, Royal Holloway, University of London matteo.fontana(a)rhul.ac.uk<mailto:matteo.fontana@rhul.ac.uk>

1 0

Seminario PRISMA lun 3 giu 2024
by Francesco Caravenna 27 May '24

27 May '24

Seminari on-line del gruppo UMI - PRISMA (http://www.umi-prisma.polito.it/ <http://www.umi-prisma.polito.it/>) I seminari PRISMA hanno un formato di "colloquium" per creare un'occasione di scambio e discussione con tutta la comunità dei probabilisti e statistici italiani. Ogni giornata comprende due relatori che tengono due seminari di 30 minuti strettamente connessi, per presentare alla comunità una prospettiva sul proprio ambito di ricerca. Da quest'anno le registrazioni dei seminari vengono pubblicate sul canale YouTube dell'UMI: https://youtube.com/playlist?list=PLmySpc-jrtAMq84VH71evyqPc1hl6eEQb <https://youtube.com/playlist?list=PLmySpc-jrtAMq84VH71evyqPc1hl6eEQb> Il prossimo appuntamento è per lunedì 3 giugno 2024. I relatori saranno Alessandro De Gregorio (Sapienza Università di Roma) e Stefano Iacus (Harvard University) che parleranno di Stimatori regolarizzati per equazioni differenziali stocastiche osservate a tempi discreti con il seguente orario: 16:00 Primo seminario 16:30 Pausa e discussione 16:45 Secondo seminario 17:15 Conclusione e discussione Trovate di seguito il riassunto. I seminari verranno trasmessi via Zoom al seguente link: https://uniroma1.zoom.us/j/82939128330 <https://uniroma1.zoom.us/j/82939128330> Meeting ID: 829 3912 8330 Vi aspettiamo numerosi! Valentina Cammarota e Francesco Caravenna %%%%%%%%%%%%%%%%%%%%%%%%%%%% RELATORI: Alessandro De Gregorio (Sapienza Università di Roma) e Stefano Iacus (Harvard University) TITOLO: Stimatori regolarizzati per equazioni differenziali stocastiche osservate a tempi discreti RIASSUNTO: Gli stimatori regolarizzati, cioè quegli stimatori che presentano dei termini di penalizzazione nella funzione di perdita, rappresentano uno strumento fondamentale nell'ambito della moderna teoria dell'apprendimento statistico. In questo seminario discuteremo problemi di stima parametrica penalizzata per equazioni differenziali stocastiche osservate a tempi discreti. Tale tema di ricerca è di recente interesse nell'ambito della statistica per processi stocastici. Nella prima parte del seminario, dopo una breve panoramica sulle tecniche di stima per equazioni differenziali stocastiche, introdurremo i modelli stocastici sparsi; ovvero si ipotizza che solo un piccolo numero di parametri determini il "vero" modello. In questo contesto è cruciale considerare degli stimatori regolarizzati che consentano di effettuare la stima e contemporaneamente la selezione del processo di diffusione. In particolare, saranno introdotti gli stimatori LASSO ed Elastic-Net e verranno discusse le loro proprietà asintotiche. La seconda parte dell'intervento sarà dedicata alle equazioni differenziali stocastiche su reti, in cui ciascun nodo della rete è un un'equazione stocastica che dipende dai nodi vicini. Questi modelli vengono introdotti poiché consentono di analizzare serie storiche ad altissima dimensione, sfruttando la struttura sparsa del grafo. Anche in questo contesto saranno introdotti e studiati stimatori penalizzati per la stima dei parametri del modello. Inoltre, la performance degli stimatori mediante sarà analizzate tramite alcune applicazioni. %%%%%%%%%%%%%%%%%%%%%%%%%%%%

1 0

Fwd: Deadline extended for ESOBE 2024
by Roberto CASARIN 27 May '24

27 May '24

fyi ---------- Forwarded message --------- Da: Sune Karlsson <Sune.Karlsson(a)oru.se> Date: lun 27 mag 2024 alle ore 14:27 Subject: Deadline extended for ESOBE 2024 To: <ECONOMETRIC-RESEARCH(a)jiscmail.ac.uk> The deadline for submitting papers to ESOBE 2024 (European Seminar on Bayesian Econometrics) has been extended to June 9, 2024. Submit a paper or an extended abstract by sending an e-mail to ESOBE2024(a)oru.se We welcome submissions from all fields of applied Bayesian econometrics (macro, micro and financial), as well as papers on statistical methodology, machine learning and computing. The conference will feature both plenary and poster sessions. The conference will take place on August 22-24 at Örebro University. Please see the conference website https://www.oru.se/esobe2024 for additional information. --- Sune Karlsson Professor of Statistics Handelshögskolan/Örebro University School of Business Phone: +46 19 301257 https://www.oru.se/hh/sune_karlsson https://econpapers.repec.org/RAS/pka1.htm E-mailing Örebro University will result in the university processing your personal data, for more information on how this is done, see https://www.oru.se/english/about-us/processing-of-personal-data-at-orebro-u…. E-mail correspondence with Örebro University may be classified as official and public documents and are handled according to archive regulations. ######################################################################## To unsubscribe from the ECONOMETRIC-RESEARCH list, click the following link: https://www.jiscmail.ac.uk/cgi-bin/WA-JISC.exe?SUBED1=ECONOMETRIC-RESEARCH&… This message was issued to members of www.jiscmail.ac.uk/ECONOMETRIC-RESEARCH, a mailing list hosted by www.jiscmail.ac.uk, terms & conditions are available at https://www.jiscmail.ac.uk/policyandsecurity/ -- Roberto Casarin, PhD Professor of Econometrics Ca' Foscari University of Venice San Giobbe 873/b - 30121 Venezia, Italy http://sites.google.com/view/robertocasarin/ https://www.unive.it/vera <https://www.unive.it/isba2024> https://www.unive.it/isba2024

1 0

Fwd: [owps] Reminder next OWPS
by Faggionato, Alessandra 27 May '24

27 May '24

Buongiorno ricevo e con piacere inoltro. Saluti Alessandra ---------- Forwarded message --------- From: One World Probability <ow.probability(a)gmail.com> Date: Mon, 27 May 2024 at 08:40 Subject: [owps] Reminder next OWPS To: <owps(a)lists.bath.ac.uk> This is a gentle reminder that the next OWPS will be *today* from *15:00* to *17:00* *CEST *time. There will be two talks on the topic of exploration driven analysis of random graphs. The talks will touch upon general techniques and specific applications of these techniques. Title, abstract and the zoom link are below the signature and can be found on the website https://www.owprobability.org/one-world-probability-seminar <https://protect-eu.mimecast.com/s/-zGkCWqjZFlpkVlsnEyR_?domain=eur01.safeli…> . ----------------------- Talk 1 :* Mariana Olvera-Cravioto, University of North Carolina at Chapel Hill* Local limits and preferential attachment graphs This talk is meant to provide an overview of local weak convergence techniques for a large class of directed random graphs, including static models such as the Erdos-Renyi, Chung-Lu, stochastic block model, and configuration models, as well as dynamic models such as the collapsed branching process and the directed preferential attachment graph. We explain how local limits can be used to study important structural graph properties, including centrality measures such as the PageRank distribution. We further use the insights obtained from our analysis of centrality measures to explain how static models and evolving models differ in how large degree vertices are distributed within the graph and how they shape their neighborhoods. In particular, we show that the empirically accepted “power-law hypothesis” on scale-free graphs, which states that the PageRank distribution follows a power-law with the same tail index as the in-degree distribution, holds in most static models but not in the dynamic models we study. This is joint work with: Sayan Banerjee and Prabhanka Deka. Talk 2 : Sayan Banerjee, *University of North Carolina at Chapel Hill* Exploration-driven networks We propose and investigate a class of random networks where incoming vertices locally explore the graph before attaching to an existing vertex. Specific instances of these networks correspond to uniform attachment, linear preferential attachment and attachment with probability proportional to vertex PageRanks. We obtain local weak limits for such networks and use them to derive asymptotics for the limiting empirical degree and PageRank distribution. We also quantify asymptotics for the degree and PageRank of fixed vertices, including the root, and the height of the network. Two distinct regimes are seen to emerge, based on the expected exploration distance of incoming vertices, which we call the ‘fringe’ and ‘non-fringe’ regimes. These regimes are shown to exhibit different qualitative and quantitative properties. In particular, networks in the non-fringe regime undergo ‘condensation’ where the root degree grows at the same rate as the network size. Networks in the fringe regime do not exhibit condensation. Interesting phase transition phenomena are exhibited for the height of the tree and the limiting PageRank distribution. The latter connects to the well-known power-law hypothesis and the proposed class of models `interpolate’ between the PageRank behavior of static and dynamic graphs discussed in Mariana’s talk. Based on joint works with Mariana Olvera-Cravioto, Shankar Bhamidi and Xiangying (Zoe) Huang. Zoom-link: https://zoom.us/j/3766827761 Meeting ID: 376 682 7761 Passcode: sPNKq1 -- ************************************************* Prof. Alessandra Faggionato https://www1.mat.uniroma1.it/people/faggionato/ Department of Mathematics University "La Sapienza" Piazzale Aldo Moro, 5 00185 - Rome Office 123, first floor *************************************************

1 0

talk Prof. Domenico Marinucci on May 27 - DIMA - UNIGE
by Ernesto De Vito 24 May '24

24 May '24

Dear all, this is a kind reminder of the following seminar Speaker : Domenico Marinucci (https://sites.google.com/view/domenicomarinucci/home) Affiliation: Dipartimento di Matematica - Università Roma Tor Vergata Title: Spectral complexity of deep neural networks Date: Monday, May 27, 2024 at 11.30 Place: Aula 704 (7th floor) , Dipartimento di Matematica at University of Genova, Via Dodencaneso 35, Teams link: https://teams.microsoft.com/l/meetup-join/19%3ameeting_NTEyMmU4MWEtZjIwOC00… Abstract: It is well-known that randomly initialized, push-forward, fully-connected neural networks weakly converge to isotropic Gaussian processes, in the limit where the width of all layers goes to infinity. In this paper, we propose to use the angular power spectrum of the limiting fields to characterize the complexity of the network architecture. In particular, we define sequences of random variables associated with the angular power spectrum, and provide a full characterization of the network complexity in terms of the asymptotic distribution of these sequences as the depth diverges. On this basis, we classify neural networks as low-disorder, sparse, or high-disorder; we show how this classification highlights a number of distinct features for standard activation functions, and in particular, sparsity properties of ReLU networks. Our theoretical results are also validated by numerical simulations. Joint work with Simmaco Di Lillo, Michele Salvi e Stefano Vigogna Best wishes, Ernesto --------------------- Ernesto De Vito DIMA - Dipartimento di Matematica MaLGa - Machine Learning Genoa center Via Dodecaneso 35 16146 Genova Italy e-mail: ernesto.devito(a)unige.it tel: +390103536783

1 0

talk Dott. Marco Tarsia on May 29, 2024
by Mastrogiacomo Elisa 24 May '24

24 May '24

Dear all, you are invited to participate to the following seminar on Wednesday 29th of May 2024 at 12:00: Title: STOCHASTIC ORDERINGS FOR SET-VALUED RISK MEASURES Abstract: We introduce and analyze from an axiomatic point of view an extension to the set-valued scenario of the maximal correlation risk measure as defined by Burgert and Rüschendorf (2006). Furthermore, we present the notion of stochastic ordering for random vectors, utilizing the upper expectation operator - introduced in Hamel and Heyde (2021) - in conjunction with extensive classes of multidimensional functions. We then explore the consistency of such stochastic orderings for appropriate set-valued risk measures. These measures resemble the maximal correlation risk measure, offering flexibility regarding adherence to all the axioms of a portfolio aggregator and a proper set-valued risk measure. A remarkable example within this category is represented by law-invariant set-valued risk measures, such as the set-valued distortion risk measure as defined, for example, in Chen and Hu (2019). The in-person presentation will be held in Room 3 Padiglione Morselli, Via Ottorino Rossi, 9, 21100 Varese VA<https://www.google.com/maps/place//data=!4m2!3m1!1s0x478681f7697827db:0x771…>, Insubria University. Scan the QR code in the attached file to join the seminar online in MS Teams. Best regards, Elisa Mastrogiacomo Elisa Mastrogiacomo ----------------------------------------- Professore Ordinario di Metodi matematici dell'economia e delle scienze attuariali e finanziarie DIPARTIMENTO D'ECCELLENZA 2023/2027 Università degli Studi dell'Insubria Dipartimento di Economia Via Monte Generoso, 71 – 21100 Varese tel. +39 0332/395528 web: www.uninsubria.it<http://www.uninsubria.it/> mail: elisa.mastrogiacomo(a)uninsubria.it<mailto:mario.rossi@uninsubria.it> Chiaramente Insubria! [LOGO-ATENEO-FONDO TRASPARENTE]<http://www.uninsubria.it/> [Facebook] <https://www.facebook.com/uninsubria> [Twitter] <https://twitter.com/Uni_Insubria> [Instagram] <https://instagram.com/uninsubria> [YouTube] <https://www.youtube.com/user/LabVAMultimedia>

1 0

JOB: Research Associate in Bayesian Statistics, Newcastle University
by Fabrizio Leisen 24 May '24

24 May '24

Con preghiera di diffusione. IMPORTANTE: Il candidato deve gia avere un dottorato di ricerca. Non sono considerati candidati che non hanno ancora ottenuto (o stanno per ottenere) un dottorato di ricerca. È richiesta buona conoscenza di Statistica Bayesiana e metodi computazionali. È una posizione di due anni basata a Newcastle. Il progetto sarà supervisionato da me, Prof. Cristiano Villa e Prof. Kevin Wilson. Cordiali Saluti, Fabrizio -- Fabrizio Leisen Professor of Statistics King’s College London Department of Mathematics Strand | WC2R 2LS | London JOB: Research Associate in Bayesian Statistics School of Mathematics, Statistics and Physics, Newcastle University Full advert: https://jobs.ncl.ac.uk/job/Newcastle-Research-Associate-in-Bayesian-Statist… Applications are invited from outstanding individuals for a Research Associate in Bayesian Statistics, to start from August/September 2024 for a period of two years. The project is at the forefront of Bayesian inference and aims to design a novel Bayesian methodology to implement Bayesian Additive Regression Trees (BART) models to deal with non-linear phenomena. In particular, the aim is to bridge the Loss-based methodology to the framework of BART, building on the previous works of Dr Cristiano Villa and Prof. Fabrizio Leisen on model selection, linear regression models and Gaussian graphical models. The successful candidate will work on developing prior distributions to estimate the structure of the trees and the number of trees in a BART model, and building scalable and efficient computational tools to ease the implementation of the methodology. The candidate must have a solid background in computational methods for Bayesian inference to develop efficient algorithms for the new proposed modelling framework. Within the project, the candidate will also perform theoretical modelling and data analysis, as well as disseminate the outputs through publications and presentations at scientific meetings. The successful candidate will join the growing Statistics group at Newcastle University and interact with the Statistics group at King's College London. The position is offered on a fixed term basis for two years from the start date, or tenable until the project end date, whichever is soonest. Generous funds are available for travel, training and other support. To apply, please complete the online application and attach a CV and covering letter by the deadline of 23rd June 2024. In your covering letter, please outline how you are either working towards, meet or exceed all the essential requirements for the role holder as outlined in the full advertisement, and highlight any expertise relevant to the project. For all information enquiries about the position, please contact Dr Kevin Wilson (kevin.wilson(a)newcastle.ac.uk<mailto:kevin.wilson@newcastle.ac.uk>), Prof Cristiano Villa (cristiano.villa(a)dukekunshan.edu.cn<mailto:cristiano.villa@dukekunshan.edu.cn>) or Prof. Fabrizio Leisen (fabrizio.leisen(a)kcl.ac.uk<mailto:fabrizio.leisen@kcl.ac.uk>). Kind Regards Kevin Dr Kevin Wilson Reader in Applied Statistics School of Mathematics, Statistics & Physics Newcastle University kevin.wilson(a)newcastle.ac.uk You may leave the list at any time by sending the command SIGNOFF allstat to listserv(a)jiscmail.ac.uk, leaving the subject line blank. -- Fabrizio Leisen Professor of Statistics King’s College London Department of Mathematics Strand | WC2R 2LS | London

1 0

UPDATED: Time Series Models for Financial and Environmental Risk - May 30 - Ca' Foscari University
by Roberto CASARIN 21 May '24

21 May '24

Dear Colleagues, On Thursday the 30th of May you are all warmly invited to our one-day workshop organised under the GRINS Project on *Time Series Models for Financial and Environmental Risk* Participation is free. However, for organizational purposes, you are kindly asked to register by May the 27th here https://bit.ly/form-workshop-timeseriesmodels . Ca’ Foscari University, San Giobbe Economics Campus, Meeting Room 1, Venezia *Program:* 10.00am Registration 10.45am *Welcome Remarks: Prof. Monica Billio, Ca’ Foscari University of Venice, Project Leder of the Grins Spoke 4, Sustainable Finance.* *1st Session* 11.00am *Prof.* *Giacomo Bormetti, University of Pavia*, *Measuring price impact and information content of trades in a time-varying setting.* 11.45am *Dr. Giuseppe Buccheri, University of Verona*, *Semiparametric Estimation of Volatility in the Presence of Intraday Drift Dynamics.* 12.30pm Lunch *2nd Session* 2.00pm* Dr. Dario Palumbo, Ca' Foscari University of Venice*, *An Observation-Driven Generalized Poisson Model*. 2.45pm *Dr. Enzo D'Innocenzo, University of Bologna, **Limit results for score-driven filters**.* 3.30pm Coffee break 4.00pm *Keynote speech**: Prof. Andrew Harvey, University of Cambridge. **Score-driven models: overview and recent developments.* *Abstracts*: *Prof.* *Giacomo Bormetti,*University of Pavia, Italy https://unipv.unifind.cineca.it/get/person/021642 *Measuring price impact and information content of trades in a time-varying setting* We propose a non-linear observation-driven version of the Hasbrouck (1991) model for dynamically estimating trades' market impact and information content. We find that market impact displays an intraday pattern superimposed with large fluctuations. Some of them are exogenous, and, as an example, we investigate market impact dynamics around FOMC announcements. Contrary to Hasbrouck (1991), we find that the information content of trades depends on the local liquidity level and the recent history of prices and trades. Finally, we use the model to estimate the time-varying permanent impact parameter, which allows performing a dynamic transaction cost analysis. Joint work with Francesco Campigli and Fabrizio Lillo. *Dr. Giuseppe Buccheri,*University of Verona, Italy https://sites.google.com/view/giuseppe-buccheri/home *Semiparametric Estimation of Volatility in the Presence of Intraday Drift Dynamics* In the conventional semimartingale representation of the log-price of a financial security, the volatility of daily log-returns is consistently estimated by their ex-post quadratic variation. A crucial assumption underlying this result is that the drift dynamics are negligible or pre-determined given past information. Recent empirical evidences of intraday unanticipated variations of mean returns call into doubt the validity of this assumption. We propose a semiparametric estimator of the daily return volatility incorporating the effect of stochastic intraday dynamics of the drift process. This framework allows us to test for the presence of drift dynamics in the data and to assess their impact on the daily return volatility. Our empirical analysis shows three main results: (i) there exists compelling evidence of intraday drift dynamics across various asset classes; (ii) when the drift moves significantly, our measure of volatility provides a better description of the return distribution; (iii) the component of the return variance associated with drift dynamics is non-negligible and possesses predictive power. Joint work with Giorgio Vocalelli. *Dr. Dario Palumbo,* Ca' Foscari University of Venice, Italy https://www.unive.it/data/people/22437798 *An Observation-Driven Generalized Poisson Model for Environmental Variables* Based on the Generalised Poisson (GP) conditional distribution, a new general class of observation-driven models for count data is presented and their theoretical properties are derived. The GP is a flexible distribution which allows for both under- and over-dispersion. As a special member of this class, a score-driven model version is introduced. We show that this specification is robust to the presence of outliers and can be extended to allow for time varying over-dispersion. For the estimation of the model we provide a Bayesian inference framework and an efficient posterior approximation procedure based on Markov Chain Monte Carlo. The applications on environmental variables show that the proposed model is well suited for capturing the over-dispersion feature of the data. *Dr. Enzo D'Innocenzo,* University of Bologna, Italy https://www.enzodinnocenzo.com/ *Limit results for score-driven filters* A study of the filtering approximation-theoretic properties of score-driven time series models. The new results are derived under a set of Lipschitz-type and tail conditions, from which I obtain maximal and deviation inequalities for the filtering approximation error. For the latter, I use techniques from empirical process theory. The novel proposed approach allows, for the first time, to study the asymptotic behavior of the empirical distribution function and the empirical processes of the approximated residuals. Since score-driven models encompasses the well-known generalized autoregressive conditional heteroskedasticity, it is proved that the results obtained in this paper extends those of Francq and Zakoian (2022) for this class of nonlinear models. In conclusion, I also give an application of these results to a popular score-driven time series models, namely, the Beta-t-GARCH(1,1) model. *Prof. Andrew Harvey,*University of Cambridge, UK *https://www.econ.cam.ac.uk/people/emeritus/ach34 <https://www.econ.cam.ac.uk/people/emeritus/ach34>* *Score-driven models: overview and recent developments* Score-driven <ach34(a)cam.ac.uk)Score-driven> time series models were introduced into the literature more than a decade ago. Since then progress has been made on the theory and a variety of new models have been introduced. I begin by reviewing theoretical results on stationarity and invertibility and show how these apply to moment-driven and score-driven models for dynamic location and/or scale parameters from a wide range of distributions. Issues of robustness are discussed and it is argued that quasi score-driven (QSD) models have an important role to play. The advantages of an exponential link function are stressed. In particular it is argued that (Q)SD-EGARCH models are much better than standard (moment-driven) GARCH models. I finish by showing how the score-driven approach solves two difficult problems: constructing time series models for censored observations and circular data. -- Roberto Casarin, PhD Professor of Econometrics Ca' Foscari University of Venice San Giobbe 873/b - 30121 Venezia, Italy http://sites.google.com/view/robertocasarin/ https://www.unive.it/vera <https://www.unive.it/isba2024> https://www.unive.it/isba2024

1 0

Time Series Models for Financial and Environmental Risk - May 30 - Ca' Foscari University
by Roberto CASARIN 21 May '24

21 May '24

Dear Colleagues, On Thursday the 30th of May you are all warmly invited to our one-day workshop organised under the GRINS Project on *Time Series Models for Financial and Environmental Risk* *https://www.unive.it/data/33113/3/88487 <https://www.unive.it/data/agenda/3/88487>* Participation is free. However, for organizational purposes, you are kindly asked to register by May the 27th here https://bit.ly/form-workshop-timeseriesmodels . Ca’ Foscari University, San Giobbe Economics Campus, Meeting Room 1, Venezia *Program:* 10.00am Registration 10.45am *Welcome Remarks: Prof. Monica Billio, Ca’ Foscari University of Venice, Project Leder of the Grins Spoke 4, Sustainable Finance.* *1st Session* 11.00am *Prof.* *Giacomo Bormetti, University of Pavia*, *Measuring price impact and information content of trades in a time-varying setting.* 11.45am *Dr. Giuseppe Buccheri, University of Verona*, *Semiparametric Estimation of Volatility in the Presence of Intraday Drift Dynamics.* 12.30pm Lunch *2nd Session* 2.00pm* Dr. Dario Palumbo, Ca' Foscari University of Venice*, *An Observation-Driven Generalized Poisson Model*. 2.45pm *Dr. Enzo D'Innocenzo, University of Bologna, **Limit results for score-driven filters**.* 3.30pm Coffee break 4.00pm *Keynote speech**: Prof. Andrew Harvey, University of Cambridge. **Score-driven models: overview and recent developments.* *Abstracts*: *Prof.* *Giacomo Bormetti,*University of Pavia, Italy https://unipv.unifind.cineca.it/get/person/021642 <https://warwick.ac.uk/dfirth> *Measuring price impact and information content of trades in a time-varying setting* We propose a non-linear observation-driven version of the Hasbrouck (1991) model for dynamically estimating trades' market impact and information content. We find that market impact displays an intraday pattern superimposed with large fluctuations. Some of them are exogenous, and, as an example, we investigate market impact dynamics around FOMC announcements. Contrary to Hasbrouck (1991), we find that the information content of trades depends on the local liquidity level and the recent history of prices and trades. Finally, we use the model to estimate the time-varying permanent impact parameter, which allows performing a dynamic transaction cost analysis. Joint work with Francesco Campigli and Fabrizio Lillo. *Dr. Giuseppe Buccheri,*University of Verona, Italy https://sites.google.com/view/giuseppe-buccheri/home <https://www.ikosmidis.com/> *Semiparametric Estimation of Volatility in the Presence of Intraday Drift Dynamics* In the conventional semimartingale representation of the log-price of a financial security, the volatility of daily log-returns is consistently estimated by their ex-post quadratic variation. A crucial assumption underlying this result is that the drift dynamics are negligible or pre-determined given past information. Recent empirical evidences of intraday unanticipated variations of mean returns call into doubt the validity of this assumption. We propose a semiparametric estimator of the daily return volatility incorporating the effect of stochastic intraday dynamics of the drift process. This framework allows us to test for the presence of drift dynamics in the data and to assess their impact on the daily return volatility. Our empirical analysis shows three main results: (i) there exists compelling evidence of intraday drift dynamics across various asset classes; (ii) when the drift moves significantly, our measure of volatility provides a better description of the return distribution; (iii) the component of the return variance associated with drift dynamics is non-negligible and possesses predictive power. Joint work with Giorgio Vocalelli. *Dr. Dario Palumbo,* Ca' Foscari University of Venice, Italy https://www.unive.it/data/people/22437798 *An Observation-Driven Generalized Poisson Model for Environmental Variables* Based on the Generalised Poisson (GP) conditional distribution, a new general class of observation-driven models for count data is presented and their theoretical properties are derived. The GP is a flexible distribution which allows for both under- and over-dispersion. As a special member of this class, a score-driven model version is introduced. We show that this specification is robust to the presence of outliers and can be extended to allow for time varying over-dispersion. For the estimation of the model we provide a Bayesian inference framework and an efficient posterior approximation procedure based on Markov Chain Monte Carlo. The applications on environmental variables show that the proposed model is well suited for capturing the over-dispersion feature of the data. *Dr. Enzo D'Innocenzo,* University of Bologna, Italy https://www.enzodinnocenzo.com/ <https://researchportal.uc3m.es/display/inv16189> *Limit results for score-driven filters* A study of the filtering approximation-theoretic properties of score-driven time series models. The new results are derived under a set of Lipschitz-type and tail conditions, from which I obtain maximal and deviation inequalities for the filtering approximation error. For the latter, I use techniques from empirical process theory. The novel proposed approach allows, for the first time, to study the asymptotic behavior of the empirical distribution function and the empirical processes of the approximated residuals. Since score-driven models encompasses the well-known generalized autoregressive conditional heteroskedasticity, it is proved that the results obtained in this paper extends those of Francq and Zakoian (2022) for this class of nonlinear models. In conclusion, I also give an application of these results to a popular score-driven time series models, namely, the Beta-t-GARCH(1,1) model. *Prof. Andrew Harvey,*University of Cambridge, UK *https://www.econ.cam.ac.uk/people/emeritus/ach34 <https://www.econ.cam.ac.uk/people/emeritus/ach34>* *Score-driven models: overview and recent developments* Score-driven <ach34(a)cam.ac.uk)Score-driven> time series models were introduced into the literature more than a decade ago. Since then progress has been made on the theory and a variety of new models have been introduced. I begin by reviewing theoretical results on stationarity and invertibility and show how these apply to moment-driven and score-driven models for dynamic location and/or scale parameters from a wide range of distributions. Issues of robustness are discussed and it is argued that quasi score-driven (QSD) models have an important role to play. The advantages of an exponential link function are stressed. In particular it is argued that (Q)SD-EGARCH models are much better than standard (moment-driven) GARCH models. I finish by showing how the score-driven approach solves two difficult problems: constructing time series models for censored observations and circular data. -- Roberto Casarin, PhD Professor of Econometrics Ca' Foscari University of Venice San Giobbe 873/b - 30121 Venezia, Italy http://sites.google.com/view/robertocasarin/ https://www.unive.it/vera <https://www.unive.it/isba2024> https://www.unive.it/isba2024

1 0

2024

2023

2022

2021

2020

2019

2018

2017

2016

2015

2014

2013

Random May 2024