WS 23/24

Course password: TBA

The course will begin on November 3rd. 

More info coming soon!

Target Audience:

-          Statistics bachelor students (preferably semester 5 and upwards) and

-          Statistics master students (credits can be counted for “Methodology and Modelling” and “Biostatistics” specification )

In this seminar, we take a closer look at areal data, spatial point processes and related topics. The students are asked to present a publication in this area. Furthermore, they have to implement the in the publication presented algorithms and models based on a further spatial data set.

Spatial point processes are stochastic processes on a subspace of multidimensional real numbers. Their realisations are finite, random and countable observations. In other words, they are point events that occur in space/on earth/etc.. Examples of such data are locations of crimes, locations of disease outbreaks or tree locations.

Areal data is a data format in which point observations are aggregated over subregions of a predefined space. These subregions are non-overlapping and make up the entire space. This format is predominately common in medical research and one of the central formats in which, for example, data on Corona infections and hospitalizations were available. Due to the loss of information on the specific point of observation, the estimation of the spatial correlation can become a bit trickier than in spatial point processes.

• The enrolment key is: "StatMetEpi2324"

Die Veranstaltung "Programmieren mit Statistischer Software (R)" wendet sich an Studenten im Bachelor Statistik (3. Semester). Sie baut auf die Veranstaltung "Einführung in die Statistische Software" (1. Semester) auf.

Official statistical agencies are the central information service providers in a democratically organized society, informing politics, business and society about current economic, social and increasingly also ecological developments. In this way, official statistics form an important basis for informed decisions. At the same time, official statistics act as a control authority, especially for politics, by empirically reflecting the consequences of decisions and actions. To meet this responsibility, official statistics production underlies a strict methodology and high-quality standards. 

Against this background, there has also been an intensive discussion about the opportunities and challenges complex statistical modelling, machine learning techniques, and new data sources deliver. This development has also led to close cooperation of statistical authorities with a number of universities, resulting in an EU-wide certification of particular master's degree programmes (EMOS: European Master in Official Statistics). For students at LMU, it is possible to obtain an EMOS certificate by taking a specific route within the Machine Learning track and the Social Statistics and Data Science track.

The course "EMOS A" aims to prepare students for participation in the discourse between academic and official statistics and the numerous opportunities for cooperation with public data producers. It provides an insight into typical questions and modes of argumentation, methodological principles, and the most important products of official statistics. First, we take a look at current research concerning official statistics. Then, the basic principles and legal framework of official statistics, their fundamental structure in Germany/Europe and their most important supporting institutions are discussed as a basis. Further topics will then include the EU system of environmental indicators, the role of Big Data in official statistics, selected aspects of result dissemination and statistical literacy, and access to official data sources for (own) secondary analyses, including an introduction to statistical anonymization techniques.

The course format is a mix of classical lecture style, guest lectures and inverted classroom elements with in-person discussion and deepening of the topics. The course is compulsory for the EMOS variant; all other students can choose it as an elective module.

Enrollment Key: emosa

Einschreibeschlüssel: sample2023

This module discusses different types of latent structures and their statistical handling. Part A, taking place in the winter semester, will focus on statistical aspects of measurement and modelling issues in surveys and assessment studies.  Part B, scheduled for the summer semester, is planned to address statistical aspects of combining data sets and different methods in the context of anonymization/privacy protection.

Concretely, Part A starts with classical testing theory (CTT) as a framework to describe the operationalization/measurement of continuous latent traits and utilizes CTT to derive and discuss critically well-known reliability measures such as Cronbach’s alpha. Then, different generalizations are studied, including structural equation models, Rasch-type models from probabilistic testing theory and situations with locally varying scales of measurement. In the second half of the semester, methods for handling incomplete data in regression models are investigated in more detail. Advanced frequentist and Bayesian correction methods for measurement error, misclassification, and missing data are developed. In this context, also an introduction to the framework of partial identification is given.

The module can be attended already in the first Master's semester.  According to the Examination Regulations, the module is among the "narrow electives" in the Social Statistics and Social Data Science track, the "wide electives" in the Econometrics and Methodology and Modelling tracks and a potential "general elective" for all tracks. Alternatively, Part A or Part B can be recognized as 'Selected Topics of Social Statistics and Social Data Science'.

Time and Dates: Tuesday, 6.15 pm to 7.45 pm, E 216 (main building)

Enrolment Key: MeasMod

Description

A variety of digital data sources are providing new avenues for empirical social science research. With the emergence of Big Data, especially data from web sources play an increasingly important role in scientific research. In order to effectively utilize these data for answering substantive research questions, a modern methodological toolkit paired with a critical perspective on data quality is needed. This course introduces computational techniques that are suited for collecting and analyzing digital behavioral data, and for exploring, visualizing and finding patterns in (different types of) data from various sources. In addition, aspects of reproducibility, data quality and error frameworks for digital data are discussed.

Enrolment Key

CSSws2324

Enrolment key: w2324-optim

Kickoff

Thu, October 19, 2pm at room S 006 (Schellingstr. 3)

Credits

6 ECTS

Format

3 hours lecture, 1 hour exercise

Description

The course introduces the theoretical foundation of optimization as well as the most prominent methods in this field. It covers the taxonomy of optimization problems and other basic principles of optimization, considering univariate and multivariate problems and commonly used approaches to tackle these. This contains first- and second-order methods as well as stochastic approaches. The course further deals with constrained optimization problems, derivative-free methods as well as multi-criteria optimization.

Contents

• Mathematical concepts
• Optimization problems
• Univariate optimization
• First order methods
• Second order methods
• Constrained optimization
• Derivative-free optimization
• Evolutionary optimization
• Bayesian optimization

Schedule

• The enrollment key is "stat_inf_w2324"

Enrolment key: InfStat12024!

Einschreibeschlüssel

• Der Einschreibeschlüssel lautet: "stat3nf2023"

Erster Termin: Voraussichtlich Donnerstag der 02. November 2023 sowohl von 10:15-11:45 Uhr: Alte Bibliothek (Raum 245, Ludwigstr. 33, 2. Stock)

, als auch von  16:15 -18:00 Uhr : Prof.-Huber-Pl. 2 (V) - LEHRTURM-V005

(Im späteren Verlauf kann dann vermutlich nur ein zeitslot angeboten werden.)

Einschreibeschlüssel: Fragestunde

Sie können sich in den Kurs mit dem Schlüssel DHS2425 selbst einschreiben.

Schedule: 

- Lecture: Monday, 16-18 c.t., Geschw.-Scholl-Pl. 1 - A 017 (starting Monday, 16 October 2023)

- Lecture/Exercise: Wednesday, 10-12 c.t., Geschw.-Scholl-Pl. 1 - A 016 (starting Wednesday, 25 October 2023)

Enrollment key: 

life23Time

Selbsteinschreibungsschlüssel: grlgprkt

Die Veranstaltung wendet sich an Studierende im Bachelor Statistik (3. Semester). Das "Grundlegende Praxisprojekt" (BA Statistik und Data Science - PO 2021) ist eine Pflichtveranstaltung (Modul P 11.1).

Die Veranstaltung wird sowohl während der Vorlesungszeit als auch in den Semesterferien angeboten. Diese Moodle-Seite ist für beide Veranstaltungen. 
Für beide Blöcke findet eine Einführungsveranstaltung mit Anwesenheitspflicht am 19.10. 10-12 statt.
Aus organisatorischen Gründen ist eine frühzeitige, separate Anmeldung für die Teilnahme während der Vorlesungszeit nötig -- schreiben Sie sich bitte ein und gehen Sie dann zur Anmeldung auf der Kursseite.

Format

Class

• Time: Thursday 10:15-11:45 h
  
• Location: M 010

Enrollment key

• The enrollment key is I2ML.

The lecture deals with theoretical and practical concepts from the fields of statistical learning and machine learning. The main focus is on predictive modeling / supervised learning. The tutorial applies these concepts and methods to real examples for illustration purposes.

key: SL_ws2324

Teilnehmende dieses Seminars werden basierend auf der offiziellen Seminareinteilung von Dr. Schollmeyer dem Moodle-Kurs hinzugefügt.

Bei Fragen wenden Sie sich bitte an ludwig.bothmann@stat.uni-muenchen.de

Master course 

• Master Statistics & Data Science – WP44 Advanced Statistical Modelling
• ESG Data Science – Elective
• Master Statistik (mit WiSo), Biostatistik – Ausgewählte Gebiete... oder Schätzen und Testen II (mit Zusatzleistung)
  

Bayesian methods have seen increasing use within statistics in the last couple of years. In Bayesian frameworks, prior knowledge can be integrated as prior distribution. These also allows to regularise parameters in large p>>n models via priors, up to the point where the prior becomes part of the statistical model itself.

In this course, advanced Bayesian models will be discussed. These include several types of latent models, in particular latent Gaussian fields, applied for example to time series and spatial data, mixed effect models and other structured regression models. The course will also cover different type of algorithms to assess the posterior, based on Monte Carlo methods and on approximation. Further topics include Bayesian model and variable selection and Bayesian classification.

Requirements: 

Students should be familiar with basics of Bayesian statistics.

Recommended: 

• Statistical inference
• Statistical modelling

Learning outcomes:

After successfully completing the course, students will be able 

• to understand Bayesian models 
• to apply Bayesian modelling for appropriate data
• to interpret the results of Bayesian models 
• to implement Bayesian models in standard software

Syllabus

Teacher: Walter J. Radermacher
Runtime: 1. October 2023 - 12. October 2023
Format: Self-learning through online videos in the first week and in-person workshops for practical appliance of use cases in the second week.
Examination: Oral Exam (3 ECTS credits)
Language: English


Who is this course for? MsC Statistics and Data Science (2021, WP 28+29+40+46), BsC Statistics and Data Science (2021, WP 8+11), Statistik und Data Science als Nebenfach für Bachelor 30 ECTS (2021, WP 4+5), Statistik und Data Science als Nebenfach für Bachelor 60 ECTS (2021, WP 11+12), Statistik und Data Science als Nebenfach 30 ECTS Mathematik (2021, WP 5+6), Statistik Nebenfach 60 ECTS Bachelor Soziologie (2021, WP 10+11), WISO Mater (2010, "Ausgewählte Gebiete..." (3ECTS))

Schedule

• Lecture: Wednesday, 9 - 11 c.t.
• Tutorial: Wednesday, 11 - 12 c.t.
• Location: Oettingenstr. 67, B U101 (Raumfinder)

Enrollment key

• Enrollment key: flash_attention