**(NMST611) Advanced Statistical Seminar (2023)**

**Wednesday: 15:40 - 17:20**| Prezenčne v Praktiku KPMS

**Aktuálne**

- Seminár začne (predpokládame a dúfame) hneď v prvom týždni semestru, štandardne v stredu (t.j., 15.02.2023) od 15:40 v Praktiku KPMS.

**Rozvrh / Seminar schedule**

**15.02.2023 | 15:40 | Patricia Martinkova**

*Institute of Computer Science, Czech Academy of Sciences, Czech Republic*

**Title: Statistical methods for analysing latent variables***In social sciences, the variable of interest is often hidden and can only be inferred indirectly through other directly measured variables. We discuss challenges for statistical analysis of such latent variables and offer some innovations in statistical methodology for measurement evaluation. We first focus on measurement error, which is omnipresent in this area, and we propose a flexible approach for assessing reliability in cases of heterogeneity due to covariates. We then discuss methods for analysing differences in academic achievement and growth.*
-------------------------------------------------------------------------------------------- 01.03.2023 | 15:40 | Canceled -------------------------------------------------------------------------------------------
**15.03.2023 | 15:40 | Kata Vuk**

*Fakultät für Mathematik, Ruhr-Universität Bochum, Germany*

**Title: Nonparametric tests based on weighted two-sample U-Statistic for change-points in time series***We investigate the limit distribution of weighted test statistics based on two-sample U-statistics. We dervie the limit distribution both under the hypothesis of no change and under the alternative of a change in mean. We consider a local alternative in which the jump height decreases as the sample size increases, but also an other type of alternative in which the time of change moves closer to the border of the observation range.*
-------------------------------------------------------------------------------------------**29.03.2023 | 15:40 | Shakeel Gavioli-Akilagun**

*Department of Statistics, London School of Economics and Political Science, United Kingdom*

**Title: Inference for Change Points in Piecewise Polynomials***We consider the problem of uncertainty quantification in change point regressions, where the signal can be piecewise polynomial of arbitrary but fixed degree. That is we seek disjoint intervals which, uniformly at a chosen confidence level, must each contain a change point location. The key theoretical tool is to perform many local tests at a range of scales and locations, and to establish weak convergence of the supremum of local test statistics to an extreme value distribution underthe null of no change points. We will discuss two algorithms for the task: the first algorithm runs in quasi-linear time and is well suited to Gaussian or sub-Gaussian data, the second algorithm is slower but makes almost no assumptions on the distribution of the data. We will discuss some appealing theoretical properties of these algorithms, and show their good practical performance on real and simulated data.*
-------------------------------------------------------------------------------------------**12.04.2023 | 15:40 | Bojana Milošević**

*aculty of Mathematics, University of Belgrade, Republic of Serbia*

**Title: Nonparametric tests: different approaches for incomplete data settings***Very often we are facing the problem of applying some statistical inferential procedure on incomplete datasets. For example, in survival analysis, the focus of the study is usually on the time until a certain event happens. The final goal of the study is to answer specific questions related to survival time. Since the studies are time-limited and a person might leave the study for some reason, the problem of having randomly right-censored data is quite common. In this talk, we overview the problem of missing data and, based on the type of missingness mechanism, present some solutions when the final goal is to test goodness- of-fit with some distributions or related hypotheses.*
-------------------------------------------------------------------------------------------**26.04.2023 | 15:40 | Siegfried Hörmann**

*Graz University of Technology, Austria*

**Title: Large sample distribution for fully functional periodicity tests***Periodicity is one of the most important characteristics of time series, and tests for periodicity go back to the very origins of the field. We consider the two situations where the potential period of a functional time series (FTS) is known and where it is unknown. For both problems we develop fully functional tests and work out the asymptotic distributions. When the period is known we allow for dependent noise and show that our test statistic is equivalent to the functional ANOVA statistic. The limiting distribution has an interesting form and can be written as a sum of independent hypoexponential variables whose parameters are eigenvalues of the spectral density operator of the FTS. When the period is unknown our test statistic is based on the maximal norm of the functional periodogram over fundamental frequencies. The limiting distribution of this object is rather delicate: it requires a central limit theorem for vectors of functional data, where the number of components increases proportional to the sample size. The talk is based on joint work with Piotr Kokoszka (Colorado State University) and Gilles Nisol (ULB) and Clément Cerovecki (ULB).*