Search
You are now searching on the website "Department of Statistics". Would you like to search the entire Stockholm University website?
Search on su.seShowing 21-40 of 110 hits
- 10 reasons to join the student council!
Department of Statistics
Our alumnus Lisa Nikiforova tells us about the benefits of joining the student council. Today Lisa has a career as a statistician and data scientist.
- Research
Department of Statistics
Here you will find information about our research, our researchers and their publications.
- Study with us
Department of Statistics
When you study statistics you learn about methods for collecting, processing, analyzing and drawing conclusions from data. You will enhance your skills in statistical programming and many of our alumni work as data scientists after their statistics studies. The department’s research areas include data science, machine learning, experimental design and official statistics. When you study with us you will develop in demand skills for your future career! Statistics is essential in social planning as well as in almost all empirical sciences. Statistics is also used in industry when developing new products, production planning, and controlling quality of produced items.
- Research subjects
Department of Statistics
Statistics is the science of methods for extracting knowledge from data. Research in statistics develops strategies for efficient data collection, flexible probability models for analysing varying data types and complex relationships, and how the associated uncertainty should be quantified and communicated.
- Conferences and seminars
Department of Statistics
At the Department of Statistics, both own and invited researchers hold seminars. Events at our department Our seminars and other events Contact
- Contact
Department of Statistics
Text
- Organisation
Department of Statistics
The Department of Statistics has around 40 employees, who are researchers, teachers and technical and administrative staff. The Department is led by a Head of department. The management also consists of a Deputy head of department, a Director of undergraduate and master´s studies, a Director of PhD studies and a Head of administration. Head of department Michael Carlson Senior lecturer Deputy head of department Jessica Franzén Senior lecturer Director of undergraduate and master´s studies Jessica Franzén Senior lecturer Director of PhD studies Tatjana von Rosen Professor Head of administration Richard Hager The Department board The Department board is the highest decision-making body at the Department of Statistics. The Board is chaired by the Head of department and includes elected representatives. The Board meets 4-5 times per semester.
- Examination at another location - student and third party guidelines and checklist
Department of Statistics
If a student is unable to attend an examination, the examination can under certain conditions be carried out elsewhere in Sweden or abroad. Great restrictiveness must be observed regarding the examination at another location and allowed only on exceptional grounds. The examination shall take place at an official agency, office or similar (e.g. a college, embassy, consulate, police station, hospital). The same conditions should apply as to the examination at Stockholm University (SU). This includes that the student should be examined at the same time and should be given the same amount of time for the exam as the students taking the exam at SU, with secluded seating without risk of interference, be entitled to the same facilities as the other students and that an invigilator (guard) will monitor the examination. The examiner decides if the exam can be given at another location and is responsible for securing that the exam can be done in a safe manner. The examiner determines in consultation with the director of undergraduate studies whether the requirements for examination at another location are met and if the department has the practical ability to organize such an occasion. It is up to the student to ensure that the requirements for examination at another location can be catered for as well as finding a local and agency official. Any costs involved must be paid by the student. Checklist The agency must appoint an officer of the agency who can act as contact liaison with us. An e-mail address and telephone number(s) are required. No further action will be undertaken by us until such information has been provided. The information shall be sent to studierektorga@stat.su.se . The agency must issue a written statement in which they declare that they will comply with the following: A secluded room will be prepared within the premises of the agency or in some equivalent locale where the examinee can be seated throughout the duration of the exam. At least one invigilator provided by the agency will be present in the room at all times during the duration of the exam. The invigilator is responsible for monitoring and securing that the rules and conditions for the exam are upheld. The general Stockholm University rules for written exams may be found at https://www.su.se/staff/organisation-governance/governing-documents-rules-and-regulations/education/regulations-for-written-seated-examination-1.516584 ; the specific rules for the exam in question will be forwarded as soon as an official liaison has been appointed and notified. All costs involved will be paid for either by the student or by the agency; Stockholm University will not be held accountable for any costs involved. The written statement shall: be provided with an official letterhead and signed by an officer of the agency, verify the name of the liaison official and include contact information, be written in English or Swedish, be sent by regular mail and received by us no later than two weeks before the exam takes place. In order to avoid unnecessary delays we recommend that an electronically readable copy of the statement (scanned or photographed) is sent by e-mail to studierektorga@stat.su.se as soon as it is issued. When the required statement has been received the examiner will determine in consultation with the director of undergraduate studies whether the requirements for examination can be met under the circumstances and will give the final decision. The student and the agency will be informed of the decision as soon as possible. If the examiner decides in favour of the application, the student and agency will be given further instructions about the exam in question. When the exam is completed and handed in by the student the agency shall: issue an officially signed statement , co-signed by the invigilator(s), stating that the exam was conducted in a proper manner and in accordance with the rules and that at least one invigilator was present at all times during the exam. The statement must be written in English or Swedish. urgently scan or photograph the pages handed in by the student and the final statement and send them in a electronically readable format by e-mail to a designated e-mail address; if the e-mail has not been received within 1 hour after the exam has ended, the student’s examination will be void. urgently send us a hard copy of the final statement, the original pages handed in by the student and other specified documents by registered mail ; if the original documents have not been received within 2 weeks after the exam date, the student’s examination will be void.
- Seminar: Hoang Nguyen, Department of Management and Engineering, Linköping University
Department of Statistics
Structured factor copulas for modelling the systemic risk of European and United States banks Seminar Date: Wednesday 8 May 2024 Time: 13.00 – 14.00 Location: Campus Albano, Lecture room 27, house 4, level 2 Factor copula models have been recently proposed for describing the joint distribution of a large number of variables in terms of a few common latent factors. In this talk, we supplement the factor copula model with a combination of a factor copula model at the first tree level and a truncated vine copula structure at a higher tree level. The model can capture different behaviours at the tail of the distribution but also remains parsimonious with interpretable economic meanings. A Bayesian procedure is employed for fast inference in multi-factor and structured factor copulas. To deal with the high-dimensional structure, a Variational Inference (VI) algorithm is applied to estimate different specifications of factor copula models. Compared to Markov Chain Monte Carlo (MCMC) sampling, the variational approximation is much faster. The bivariate copula functions connecting the variables in factor copula models are unknown in high dimensions. An automatic procedure based on the Bayesian Information Criterion (BIC) is derived to recover the hidden dependence structure. We use the structured factor copulas to model the joint and conditional distress probabilities of banks in Europe and the United States.
- Seminar: Matias Quiroz School of Mathematical and Physical Sciences, University of Technology Sydney
Department of Statistics
Dynamic linear regression models for semi long memory Seminar Date: Wednesday 5 June 2024 Time: 13.00 – 14.00 Location: Campus Albano, Lecture room 27, house 4, level 2 Abstract Dynamic linear regression models forecast the values of one time series based on a linear combination of a set of exogenous time series, while incorporating a time series process for the error term. This error process is often assumed to follow an autoregressive integrated moving average (ARIMA) model, or seasonal variants thereof, which are unable to capture a long-range dependency structure of the error process. We propose a novel dynamic linear regression model that incorporates such dependency and show that it improves the model's forecasting ability. We illustrate the model's superior predictive accuracy in two energy forecasting applications.
- Thesis defence - Oscar Oelrich
Department of Statistics
Ingress
- Seminar: Sune Karlsson, Örebro University
Department of Statistics
The relation between treasury yields and the corporate bond-yield spread -- stable or time-varying? Seminar Date: Wednesday 8 November 2023 Time: 13.00 – 14.00 Location: Campus Albano, Lecture room 28, house 4, level 2 Abstract We assess whether the relation between US Treasury yields and the corporate bond yield spread has been stable over time. This is done by estimating hybrid time-varying parameter Bayesian VAR models with stochastic volatility where we - as a methodological contribution - also allow for disturbances with heavy tails. We analyse monthly data from April 1953 to February 2023 both within- and out-of-sample. Our results indicate that the relation has not been stable; more specifically, there is evidence that the equation of the corporate bond yield spread is subject to time-variation in its parameters. We also find that an increase in the corporate bond yield spread decreases the risk free rate. Finally, we note that while allowing for heavy tails receives a fair amount of support within sample, it appears to be of more limited importance from a forecasting perspective.
- Meet us at Stockholm University Virtual Master´s Fair
Department of Statistics
Are you considering applying to our master’s programme at the Department of Statistics? Join us on 18 November 2021 during Stockholm University’s Virtual Masters Fair to learn more about our programme and get all your questions answered. Event Date: Thursday 18 November 2021 Time: 12.00 – 17.00 Location: Online
- Seminar: Ganna Fagerberg, Statistiska institutionen, SU
Department of Statistics
Time-varying multi-seasonal ARMA models Seminar Date: Wednesday 26 March 2025 Time: 13.00 – 14.00 Location: Campus Albano, lecture room 25, house 4, level 2 Abstract This seminar will present a multi-seasonal time-varying ARMA model designed to capture complex patterns in both regular and seasonal components over time. The parameter evolution is modeled using dynamic shrinkage process priors, allowing for gradual changes, abrupt shifts, and extended periods of stability in response to underlying data dynamics. The model is parameterized to ensure stability and invertibility at each time point. A Gibbs sampler is developed with a fast approximate sampler based on the extended Kalman filter. The effectiveness of the proposed methodology is demonstrated through both simulated and real-world data.
- Ganna Fagerberg, dept. of Statistics, Stockholm University
Department of Statistics
Time-Varying Multi-Seasonal ARMA Models with Global-Local Shrinkage Priors
- Seminar: Sebastian Ankargren, Spotify AB
Department of Statistics
Randomized controlled trials at scale: A/B testing at Spotify Seminar Date: Wednesday 3 May 2023 Time: 13.00 – 14.00 Location: Campus Albano, lecture room 27, house 4, level 2 Abstract At Spotify, we have hundreds of experiments, so-called A/B tests, running at any given point in time. A/B testing is a form of online controlled experimentation that builds on the foundational ideas of randomized controlled trials (RCTs). What makes A/B tests different from RCTs is the environment in which they are used and the requirements they face. A/B tests are often cheap, they can easily involve hundreds of millions of units, they can often have effects that are heterogeneous across units, and they can be risky but only in terms of degrading the user experience. In this talk, I will give a high-level overview of how we do online experimentation at Spotify and then focus on one challenge we are currently occupied with: sequential inference when outcomes are measured over open-ended windows of time.
- Seminar: Azadeh Chizarifard, Department of Statistics, Stockholm University
Department of Statistics
New Approach to Studying Equivalence of BLUEs & BLUPs in Linear Mixed Models with Different Covariance Matrices Seminar Date: Wednesday 30 August 2023 Time: 13.00 – 14.00 Location: Campus Albano, Lecture room 25, house 4, level 2 Abstract Linear mixed models (LMMs) extend fixed effects linear models by allowing both fixed and random effects. LMMs are an important class of statistical models that can be used to analyze, e.g., clustered, panel, and longitudinal data with complex structures and various sources of variations. The covariance matrix structure is of utmost importance for the inference about LMM. However, determining the correct covariance matrix can be challenging, even in some relatively simple situations. Hence, extensive research has focused on finding covariance matrices that lead to equivalent best linear unbiased estimators (BLUEs) and best linear unbiased predictors (BLUPs) for the fixed and random effects, respectively. The thesis derives necessary and sufficient conditions for obtaining equal BLUEs and BLUPs under two LMMs with different covariance structures. These conditions are derived via a system of matrix equations. It has been shown that the consistency conditions for these systems of matrix equations are connected to the existing necessary and sufficient conditions for obtaining equal BLUEs and BLUPs under two LMMs with different covariance structures derived by using different approaches.
- Seminar: Stina Zetterström, Department of Statistics, Uppsala University
Department of Statistics
Bounding the selection bias Seminar Date: Wednesday 8 March 2023 Time: 13.00 – 14.00 Location: Campus Albano, lecture room 27, house 4, level 2 Abstract Selection bias is a systematic error that can occur when subjects are included or excluded in the analysis based on some selection criteria for the study population. This type of bias can jeopardize the validity of the study, not only in the case of generalization to the total population but also for the selected study population. Thus, methods for assessing the effect of selection bias are desired. One method of estimating the effect of selection bias is through sensitivity analysis, and one type of sensitivity analysis is bounding the bias. In this work, we investigate a previously proposed bound, referred to as the SV bound, that is based on assumptions of values of sensitivity parameters. Furthermore, we derive feasible regions for the sensitivity parameters as well as conditions for the SV bound to be sharp, where sharp means that the bias can be equal to the value of the bound. It can sometimes be difficult to specify the unknown sensitivity parameters and thus, as an alternative, we present a second bound that is based solely on the observed data and is therefore referred to as the assumption-free (AF) bound. Furthermore, we develop an R package for calculating the SV and AF bounds. The theoretical properties of the bounds as well as the R package are illustrated with a simulated dataset that emulates a study on the effect of the zika virus on microcephaly in Brazil. The methods, code, and results developed in this work are of interest to researchers in observational studies.
- Seminar: Paul Dickman, Medical Epidemiology and Biostatistics, Karolinska Institutet
Department of Statistics
On the proportional hazards assumption in Cox regression Seminar Date: Wednesday 13 March 2024 Time: 13.00 – 14.00 Location: Campus Albano, Lecture room 14, house 2, level 2 I will start by giving a brief overview of the proportional hazards assumption (focusing on Cox regression) and discuss the 2020 paper by Mats Stensrud and Miguel Hernán "Why Test for Proportional Hazards?" ( https://doi.org/10.1001/jama.2020.1267 ). Their statement that ‘statistical tests for proportional hazards are unnecessary’ is potentially controversial, but I don't disagree. I am concerned that the statement may be (mis)interpreted by some as ‘assessing proportional hazards is unnecessary’.
- Seminar: Aishwarya Bhaskaran, Macquarie University, Sydney
Department of Statistics
Usable and precise asymptotics for generalized linear mixed model analysis Seminar Date: Wednesday 19 June 2024 Time: 13.00 – 14.00 Location: Campus Albano, Lecture room 27, house 4, level 2 Abstract Generalized linear mixed models are an essential group of models for analysing many present-day complex data sets, especially those that contain non-normal and correlated response data. Despite the large volume of research concerning this group of models, there is little theory concerning the statistical properties of maximum likelihood estimators. Existing theoretical results available for the asymptotic variance-covariance matrix for such estimators contain limits and expectations over the response distribution. Hence such results are not easy to implement for tasks such as confidence interval construction. We present precise asymptotic results that have explicit and simple studentizable forms for likelihood-based generalized linear mixed model analysis. We also look at finite sample performance via a simulation study.