BEGIN:VCALENDAR VERSION:2.0 PRODID:researchseminars.org CALSCALE:GREGORIAN X-WR-CALNAME:researchseminars.org BEGIN:VEVENT SUMMARY:Prof. Dr. Igor Mandel (NJ\, USA) & Prof.Dr. Stan Lipovetsky (MN\, USA) (Retired) DTSTART:20260402T040000Z DTEND:20260402T050000Z DTSTAMP:20261025T092952Z UID:AMIS/1 DESCRIPTION:Title: Sim ulation-Based Insights and Novel Criteria for Linear Regression Modeling\nby Prof. Dr. Igor Mandel (NJ\, USA) & Prof.Dr. Stan Lipovetsky (MN\, U SA) (Retired) as part of Asymptotic Methods in Statistics\n\nAbstract: TBA \n\nAbstract: We study asymptotic behavior of the averaged integrals of a Lévy-driven\nlinear process weighted by a complex exponent of polynomials with real coefficients.\nSuch functionals naturally arise in the problems relating to nonlinear regression\nanalysis and signal processing\, specif ically in the estimation of parameters of\nfrequency-modulated signals.\n Under some conditions on the Lévy process and kernel defining the linea r process\,\nwe get a uniform strong law of large numbers for this weighte d process. More\nprecisely\, it is shown that the considered integrals con verge a.s. to zero uniformly\nover all the values of the real coefficients of the polynomials of fixed order.\n The result obtained is then used t o prove strong consistency of LSE for the\nparameters of linearly-modulate d trigonometric signal (chirp signal) observed against\nthe background of shot noise described above.\n LOCATION:https://researchseminars.org/talk/AMIS/1/ END:VEVENT BEGIN:VEVENT SUMMARY:PhD student Viktor Hladun (National Technical University of Ukrain e “Igor Sikorsky Kyiv Polytechnic Institute”) (National Technical Univ ersity of Ukraine “Igor Sikorsky Kyiv Polytechnic Institute”) DTSTART:20260408T140000Z DTEND:20260408T150000Z DTSTAMP:20261025T092952Z UID:AMIS/2 DESCRIPTION:Title: On uniform Strong Law of Large Numbers for weighted shot noise and consistenc y of the Least Squares Estimator of chirp signal parameters\nby PhD st udent Viktor Hladun (National Technical University of Ukraine “Igor Siko rsky Kyiv Polytechnic Institute”) (National Technical University of Ukra ine “Igor Sikorsky Kyiv Polytechnic Institute”) as part of Asymptotic Methods in Statistics\n\n\nAbstract\nWe study asymptotic behavior of the a veraged integrals of a Lévy-driven\nlinear process weighted by a complex exponent of polynomials with real coefficients.\nSuch functionals naturall y arise in the problems relating to nonlinear regression\nanalysis and sig nal processing\, specifically in the estimation of parameters of\nfrequenc y-modulated signals.\n Under some conditions on the Lévy process and ke rnel defining the linear process\,\nwe get a uniform strong law of large n umbers for this weighted process. More\nprecisely\, it is shown that the c onsidered integrals converge a.s. to zero uniformly\nover all the values o f the real coefficients of the polynomials of fixed order.\n The result obtained is then used to prove strong consistency of LSE for the\nparamete rs of linearly-modulated trigonometric signal (chirp signal) observed agai nst\nthe background of shot noise described above.\n\nThe results are join t with Prof. Dr. Alexander Ivanov (National Technical University of Ukrain e\n“Igor Sikorsky Kyiv Polytechnic Institute”).\n LOCATION:https://researchseminars.org/talk/AMIS/2/ END:VEVENT BEGIN:VEVENT SUMMARY:Doctor of Phys. and Math. Sc.\, Leading Researcher Sergiy Shklyar (Institute of Geological Sciences NAS of Ukraine) DTSTART:20260415T140000Z DTEND:20260415T150000Z DTSTAMP:20261025T092952Z UID:AMIS/3 DESCRIPTION:Title: Mul tiframe resolution enhancement in a frequency domain\nby Doctor of Phy s. and Math. Sc.\, Leading Researcher Sergiy Shklyar (Institute of Geologi cal Sciences NAS of Ukraine) as part of Asymptotic Methods in Statistics\n \nAbstract: TBA\n LOCATION:https://researchseminars.org/talk/AMIS/3/ END:VEVENT BEGIN:VEVENT SUMMARY:Prof. Dr. Baran Sandor (University of Debrecen\, Hungary) DTSTART:20260422T140000Z DTEND:20260422T150000Z DTSTAMP:20261025T092952Z UID:AMIS/4 DESCRIPTION:Title: Fai r Scores for Multivariate Gaussian Forecasts\nby Prof. Dr. Baran Sando r (University of Debrecen\, Hungary) as part of Asymptotic Methods in Stat istics\n\nAbstract: TBA\n LOCATION:https://researchseminars.org/talk/AMIS/4/ END:VEVENT BEGIN:VEVENT SUMMARY:Ptof. Dr. Volodymyr V. Anisimov (Data Science Director\, Data Scie nce\, Center for Design & Analysis\, Amgen\, London\, UK) DTSTART:20260506T140000Z DTEND:20260506T150000Z DTSTAMP:20261025T092952Z UID:AMIS/5 DESCRIPTION:Title: Adv anced Data-Driven Statistical Technologies for Designing and Forecasting O peration in Late-stage Clinical Trials\nby Ptof. Dr. Volodymyr V. Anis imov (Data Science Director\, Data Science\, Center for Design & Analysis\ , Amgen\, London\, UK) as part of Asymptotic Methods in Statistics\n\n\nAb stract\nAbstract: Clinical trials in the modern era are characterized by t heir complexity and very high costs. With the need to recruit hundreds or even thousands of patients across multiple clinical sites and countries\, conducting efficient and effective trials has become a major challenge.\nD esigning and forecasting clinical trial operations remains one of the most pressing challenges in modern drug development\, with inefficient patient enrolment being a leading contributor to costly delays. \nThis talk prese nts recent advances in analytic and statistical methodologies aimed at imp roving the predictability and efficiency of clinical trial operation.\nWe introduce innovative data-driven technologies that are based on a rigorous and practical statistical framework (hierarchic stochastic models with ra ndom parameters) and enhance recruitment forecasting by accounting for key sources of uncertainty\, including variability in site activation timelin es\, heterogeneous enrolment rates across sites\, and temporal stochastici ty. These models enable dynamic\, stage-specific projections that better a lign operational plans with real-world trial behavior.\nA framework for op timizing cost-efficient recruitment strategies through intelligent site an d country selection is also presented. This methodology incorporates opera tional constraints such as regional enrolment caps and costs to balance fe asibility and resource allocation.\nInterim reforecasting approaches that leverage accumulating data to adaptively adjust recruitment plans are disc ussed with the goal of achieving the probability of meeting enrolment mile stones. Additionally\, statistical techniques for centralized monitoring a re introduced to identify atypical performance patterns\, flagging under- or over-performing sites and informing operational interventions.\nThe tal k also covers methods for forecasting key operational metrics critical to trial planning and oversight—such as projecting event accrual in oncolog y trials. \nThe utility of these approaches is demonstrated using various case studies that illustrate their application in complex\, global clinica l programs and show how these advanced tools are reshaping clinical trial operations\, cost management\, and ultimately improved outcomes.Collective ly\, these innovations can significantly improve trial predictability and efficiency and accelerate the drug development process.\nOur research work "Forecasting and cost-efficient designing restricted enrolment in clinica l trials" was recognized by the 2025 Award for Statistical Excellence in t he Pharmaceutical Industry from the Royal Statistical Society and PSI (UK) .\n LOCATION:https://researchseminars.org/talk/AMIS/5/ END:VEVENT BEGIN:VEVENT SUMMARY:PhD student Daniel Gorbunov (Taras Shevchenko National University of Kyiv) DTSTART:20260513T140000Z DTEND:20260513T150000Z DTSTAMP:20261025T092952Z UID:AMIS/6 DESCRIPTION:Title: Non parametric regression estimators for mixtures with varying concentrations< /a>\nby PhD student Daniel Gorbunov (Taras Shevchenko National University of Kyiv) as part of Asymptotic Methods in Statistics\n\n\nAbstract\nFinite mixture models naturally arise in statistical analysis of biological and sociological data. If the sub-population which a subject belongs to is not known exactly\, the distribution of its variables is a mixture of the sub -populations’ distributions. In the classical finite mixture models (FMM ) the concentrations of the components in the mixture (mixing probabilitie s) are the same for all observations. In a more flexible mixture with vary ing concentrations model (MVC)\, the concentrations are different for diff erent observations.\n\nRegression models are typically applied to describe dependency between different numerical variables of one subject. In the c ase of homogeneous sample there exist many non-parametric estimators of th e regression function\, such as the Nadaraya-Watson estimator (NWE) and lo cal linear regression estimator (LLRE). For homogeneous samples\, NWE demo nstrates an inappropriate bias in points where the regressor probability d ensity function (PDF) has discontinuity (jump points). For such a scenario \, the LLRE stands as a remedy\, having a significantly smaller bias.\n\nI n this talk\, we consider a modification of NWE (mNWE) and LLRE (mLLRE) fo r the estimation of the regression function of some MVC component. We will show that under suitable assumptions\, the modified estimators are asympt otically normal. Moreover\, the rate of convergence for the mNWE is differ ent at different points of continuity and discontinuity of the regressor's PDF respectively\, whereas the mLLRE preserves the same rate of convergen ce for both cases.\n LOCATION:https://researchseminars.org/talk/AMIS/6/ END:VEVENT BEGIN:VEVENT SUMMARY:Dr. Yurii Malitskyi (Assistant Professor) (Faculty of Mathematics\ , University of Vienna\, Austria) DTSTART:20260520T140000Z DTEND:20260520T150000Z DTSTAMP:20261025T092952Z UID:AMIS/7 DESCRIPTION:Title: Imp licit bias in optimization algorithms\nby Dr. Yurii Malitskyi (Assista nt Professor) (Faculty of Mathematics\, University of Vienna\, Austria) as part of Asymptotic Methods in Statistics\n\n\nAbstract\nIn this talk\, we will explore the role of implicit bias in optimization algorithms. Implic it bias refers to the tendency of an algorithm to converge to a specific s olution even in the absence of an explicit regularization term in its form ulation. In other words\, implicit bias emerges from how the algorithm its elf interacts with the objective function it’s trying to minimize. We wi ll demonstrate this concept through a few surprising applications.\n\nAbst ract. In this talk\, we will explore the role of implicit bias in optimiz ation algorithms. Implicit bias refers to the tendency of an algorithm to converge to a specific solution even in the absence of an explicit regular ization term in its formulation. In other words\, implicit bias emerges fr om how the algorithm itself interacts with the objective function it’s t rying to minimize. We will demonstrate this concept through a few surprisi ng applications.\n LOCATION:https://researchseminars.org/talk/AMIS/7/ END:VEVENT BEGIN:VEVENT SUMMARY:Prof. Dr. Iryna Rozora (Prof. Dr. Iryna Rozora (Head of Department of Applied Statistics\, Faculty of Cybernetics and Computer Sciences\, Ta ras Shevchenko National University of Kyiv) DTSTART:20260527T140000Z DTEND:20260527T150000Z DTSTAMP:20261025T092952Z UID:AMIS/8 DESCRIPTION:Title: On the Statistical Properties of Impulse Response Function Estimators\nby Prof. Dr. Iryna Rozora (Prof. Dr. Iryna Rozora (Head of Department of App lied Statistics\, Faculty of Cybernetics and Computer Sciences\, Taras She vchenko National University of Kyiv) as part of Asymptotic Methods in Stat istics\n\n\nAbstract\nEstimating stochastic linear systems using Impulse R esponse Functions (IRFs) is a fundamental problem with broad relevance acr oss multiple disciplines. IRFs describe how a system responds to external inputs\, providing essential insights into system behavior and control. Su ch analyses have important applications in signal processing\, control the ory\, econometrics\, oceanography\, and related fields.\nIn this talk\, I will focus on the estimation of IRFs for Single-Input Single-Output (SISO) systems. We consider a time-invariant continuous linear system driven by zero-mean stationary Gaussian input processes. The aim of the study is to develop a method for estimating the unknown IRF based on observations of b oth the system’s input and output signals. Our approach is based on a cr oss-correlogram estimator\, which computes the correlation between the inp ut and output processes. We establish conditions for asymptotic unbiasedne ss and consistency\, analyze the rate of convergence\, and propose a stati stical goodness-of-fit test for validating the estimated IRF. The theoreti cal results are supported by numerical simulations.\n LOCATION:https://researchseminars.org/talk/AMIS/8/ END:VEVENT BEGIN:VEVENT SUMMARY:Dr. Andriy Malenko (senior researcher of V. M. Glushkov Institute of Cybernetics NAS of Ukraine) DTSTART:20260603T140000Z DTEND:20260603T150000Z DTSTAMP:20261025T092952Z UID:AMIS/9 DESCRIPTION:Title: Ada ptive radar system configuration tuning\nby Dr. Andriy Malenko (senior researcher of V. M. Glushkov Institute of Cybernetics NAS of Ukraine) as part of Asymptotic Methods in Statistics\n\n\nAbstract\nThe talk focuses o n an algorithm for parameters tuning for radar-based sensing systems for m otion monitoring. The algorithm adapts parameters for better system perfor mance according to the estimated motion zone properties\, i.e. mean and ma ximum distance to objects and their velocities. A new approach for determi nation of radar height position and tilt angle is proposed and experimenta lly tested. The method uses radar measurement points cloud only and may be applied for optimization of distance and velocity ranges and reduction of artifacts. Error of tilt angle estimation is 0.7-1.6 degrees in range of 0-45 degrees. Height estimation error depends on moving object height and is less than 1 m. The main goal of algorithm is estimation of motion zone and corresponding adaptive change of radar-system configuration parameters for optimizing system's field of view and performance. The proposed appro ach may be used for indoor and outdoor motion detection and activity class ification in applications such as health diagnostics\, health monitoring\, surveillance\, occupancy sensing\, and automotive.\n LOCATION:https://researchseminars.org/talk/AMIS/9/ END:VEVENT BEGIN:VEVENT SUMMARY:Prof. Dr. Igor Mandel (NJ\, USA) DTSTART:20260902T150000Z DTEND:20260902T160000Z DTSTAMP:20261025T092952Z UID:AMIS/10 DESCRIPTION:Title: Hu man vs. Computational Aesthetic Evaluation in Statistical Models of Art Pr ices\nby Prof. Dr. Igor Mandel (NJ\, USA) as part of Asymptotic Method s in Statistics\n\n\nAbstract\nThis presentation provides a concise survey of the topic\, spanning from hedonic regression models developed over the last 15-20 years to contemporary models based on automatic feature extrac tion and the extensive application of artificial intelligence. Particular emphasis is placed on the comparison between human and computational asses sments of the aesthetic characteristics of paintings and their relationshi p to price. In addition to the mathematical aspects of the problem\, numer ous illustrative examples of painting evaluations will be presented.\n[1] Igor Mandel. Predicting Art Market Prices with AI-Based Aesthetic Evaluati on\, SSRN\, 2026.\n[2] Igor Mandel\, Stefano Mastandrea\, Ivan Shamshurin\ , Kirill Rachenkov. Beauty or money? Statistical analysis. Model Assisted Statistical Applications\, Volume 16\, Issue 2\, 2021.\n[3] Igor Mandel. A esthetic\, Art-Historical and Economic Values in Painting: Empirical Study \, in: Integrative Explorations of the Creative Mind\, Cambridge Scholars Publishing\, 2019.\n\nNice paintings will be shown. Prof. Dr. Igor Mandel is a well-known expert in data processing.\n LOCATION:https://researchseminars.org/talk/AMIS/10/ END:VEVENT BEGIN:VEVENT SUMMARY:Dr. Yevhen Turchyn (Oles Honchar National University of Dnipro (Uk raine)) DTSTART:20260909T140000Z DTEND:20260909T150000Z DTSTAMP:20261025T092952Z UID:AMIS/11 DESCRIPTION:Title: Th e Arctan-Tan-G Family of Distributions: Properties and Applications\nb y Dr. Yevhen Turchyn (Oles Honchar National University of Dnipro (Ukraine) ) as part of Asymptotic Methods in Statistics\n\n\nAbstract\nWe introduce a new family of continuous distributions\, a so-called \nArctan-Tan-G (or ATT-G) family. Some mathematical properties of the new family are studied including\, in particular\, asymptotics for the pdf and the cdf\,\norder s tatistics\, and stochastic ordering. A special case of this family\, the A TT-Weibull (ATT-W) distribution\, is discussed. The density function\, th e quantile function\, Bowley's skewness and Moors' kurtosis are researche d. Distribution of order statistics is obtained. The model parameters are estimated by maximum likelihood. Usefulness of the new family is proven em pirically by means of a real data set.\n\nDr. Yevhen Tiurchyn is Associate Professor of Department of Statistics & Probability Theory in Oles Honcha r National University of Dnipro (Ukraine).\n LOCATION:https://researchseminars.org/talk/AMIS/11/ END:VEVENT BEGIN:VEVENT SUMMARY:Dr. Ievhen Turchyn (Oles Honchar National University of Dnipro (Uk raine)) DTSTART:20260916T140000Z DTEND:20260916T150000Z DTSTAMP:20261025T092952Z UID:AMIS/12 DESCRIPTION:Title: Th e Arctan-Tan-G Family of Distributions: Properties and Applications\nb y Dr. Ievhen Turchyn (Oles Honchar National University of Dnipro (Ukraine) ) as part of Asymptotic Methods in Statistics\n\n\nAbstract\nWe introduce a new family of continuous distributions\, a so-called \nArctan-Tan-G (or ATT-G) family. Some mathematical properties of the new family are studied including\, in particular\, asymptotics for the pdf and the cdf\,\norder s tatistics\, and stochastic ordering. A special case of this family\, the A TT-Weibull (ATT-W) distribution\, is discussed. The density function\, th e quantile function\, Bowley's skewness and Moors' kurtosis are researche d. Distribution of order statistics is obtained. The model parameters are estimated by maximum likelihood. Usefulness of the new family is proven em pirically by means of a real data set.\n\nThe talk dated September 9 is sh ifted to September 16.\n LOCATION:https://researchseminars.org/talk/AMIS/12/ END:VEVENT BEGIN:VEVENT SUMMARY:MS student Oleksandr Liubimov (Taras Shevchenko National Universit y of Kyiv and EPFL\, Lausanne\, Switzerland) DTSTART:20260925T140000Z DTEND:20260925T150000Z DTSTAMP:20261025T092952Z UID:AMIS/13 DESCRIPTION:Title: Po lynomial regression under a mixture of classical and Berkson errors with u nknown moments of the latter error\nby MS student Oleksandr Liubimov ( Taras Shevchenko National University of Kyiv and EPFL\, Lausanne\, Switzer land) as part of Asymptotic Methods in Statistics\n\n\nAbstract\nPreviousl y we gave consistent estimators for model parameters given the first 2d mo ments of Berkson error\, where d is the power of the polynomial. Now we we aken this assumption and construct new estimators for the model parameters . In particular we estimate consistently the regression coefficients witho ut information on moments of Berkson error.\n LOCATION:https://researchseminars.org/talk/AMIS/13/ END:VEVENT BEGIN:VEVENT SUMMARY:Prof. Dmytro Koroliuk (Igor Sikorsky Kyiv Polytechnic Institute\, Faculty of Electronics) DTSTART:20260930T153000Z DTEND:20260930T163000Z DTSTAMP:20261025T092952Z UID:AMIS/14 DESCRIPTION:Title: Sp ecifics of the frequency-statistical approach in the study of quantum obje cts\nby Prof. Dmytro Koroliuk (Igor Sikorsky Kyiv Polytechnic Institut e\, Faculty of Electronics) as part of Asymptotic Methods in Statistics\n\ n\nAbstract\nA quantum state does not generally specify the outcome of an individual measurement\; it specifies the statistical structure of possibl e outcomes for a specified measurement procedure.\nUsing several quantum c oncepts as examples\, the significant difference between statistical appro aches in the quantum and classical cases is demonstrated. This illustrates who was right in the debate between Albert and Niels: does God play chess or dice?\n LOCATION:https://researchseminars.org/talk/AMIS/14/ END:VEVENT BEGIN:VEVENT SUMMARY:Prof. Dr. Nick Cogan (Department of Mathematics Florida State Uni versity USA) DTSTART:20261007T140000Z DTEND:20261007T150000Z DTSTAMP:20261025T092952Z UID:AMIS/15 DESCRIPTION:Title: Br idging Mathematical Modeling and Experimental Data to Optimize Cancer Ther apy\nby Prof. Dr. Nick Cogan (Department of Mathematics Florida State University USA) as part of Asymptotic Methods in Statistics\n\n\nAbstract \nOptimizing cancer treatment will increasingly rely on close coupling bet ween experimental research and mathematical modeling. A central obstacle i n oncology is our incomplete understanding of how therapies reshape the tu mor microenvironment and how immune activity drives tumor regression. This gap is especially acute in aggressive cancers such as triple-negative bre ast cancer (TNBC)\, which lacks standard biomarkers and has limited treatm ent options. Our collaborative work develops a quantitative framework for studying combined immune and chemotherapeutic interventions. Using a novel radiolabeled T-cell tracking approach\, we follow immunomodulation over t ime and use these measurements to guide the design of optimal therapy sche dules. In this talk\, I will present ordinary differential equation (ODE) models of combination therapy dynamics and compare model performance with and without T-cell data. I will also discuss broader directions for this w ork\, including extensions to organ-level analysis and the use of optimal control methods to refine treatment strategies. By integrating experiment and theory\, we aim to strengthen the predictive power of cancer treatment models and move toward more effective\, personalized therapeutic protocol s.\n\nhttp://www.math.fsu.edu/~cogan\n LOCATION:https://researchseminars.org/talk/AMIS/15/ END:VEVENT BEGIN:VEVENT SUMMARY:Dr. Sc. (Technics) Serhii Zabolotnii (ORCID 0000-0003-0242-2234) ( Cherkasy State Business College) DTSTART:20261014T140000Z DTEND:20261014T150000Z DTSTAMP:20261025T092952Z UID:AMIS/16 DESCRIPTION:Title: Se cond-Order Least Squares and the Polynomial Maximization Method: An Estima ting-Equation Identity and a Higher-Degree Efficiency Reserve\nby Dr. Sc. (Technics) Serhii Zabolotnii (ORCID 0000-0003-0242-2234) (Cherkasy Sta te Business College) as part of Asymptotic Methods in Statistics\n\n\nAbst ract\nFor linear regression with i.i.d. homoskedastic errors\, optimally w eighted second-order least squares (SLS) is the degree-two polynomial maxi mization method (PMM) estimating equation: both select the same optimal co mbination of the residual e and of e^2 - sigma^2\, share one influence fun ction\, and attain the slope variance c_2 g_2 / N\, where g_2 = 1 - gamma_ 3^2 / (2 + gamma_4) depends only on the skewness and kurtosis of the error law. The identity is one of estimating equations\, not merely of asymptot ic variances: the Kunchenko centered-correlation matrix F_2 is the SLS wei ghting operator after a change of variables\, and the design-dependent fac tors of the SLS residual cancel in the slope score. The identity stops whe re the optimal combination depends on the design\; outside that scope cond itional SLS is the more general object. Within the polynomial moment class \, degree three holds an efficiency reserve: on four right-skewed laws the cubic direction adds between 30% and 50% asymptotic efficiency\, and unde r symmetric platykurtic errors it is the only informative direction\, sinc e g_2 = 1 there. The reserve belongs to the degree-three span\, which an e fficient generalized method of moments estimator also attains\; the PMM no rmal system is the low-dimensional route to it\, and the difference is vis ible at n <= 200. The talk gives the asymptotic law of the feasible plug-i n estimator\, a bootstrap pretest for choosing the degree\, Lean 4 checks of the degree-specific algebra\, simulation evidence\, and a residual-cali brated study on drilling innovations from the Utah FORGE 58-32 well.\nPrep rint: arXiv:2606.11421\n LOCATION:https://researchseminars.org/talk/AMIS/16/ END:VEVENT BEGIN:VEVENT SUMMARY:Dr. Sc. (Phys. & Math.)\, Associate Prof. Kostiantyn Ralchenko (Ta ras Shevchenko National University of Kyiv and EPFL\, Lausanne\, Switzerla nd) DTSTART:20261104T150000Z DTEND:20261104T160000Z DTSTAMP:20261025T092952Z UID:AMIS/17 DESCRIPTION:Title: As ymptotic properties of drift parameter estimators in CKLS diffusion models \nby Dr. Sc. (Phys. & Math.)\, Associate Prof. Kostiantyn Ralchenko (T aras Shevchenko National University of Kyiv and EPFL\, Lausanne\, Switzerl and) as part of Asymptotic Methods in Statistics\n\nInteractive livestream : https://knu-ua.zoom.us/j/89643295643?pwd=eTBZZSt0d0thZzFyaUhDUFNGTVE3QT0 9 Passcode (if necessary) 785163\n\nAbstract\nThe Chan–Karolyi–Longst aff–Sanders (CKLS) model is a flexible family of diffusion models widely used in financial mathematics\, particularly for modelling interest rates and stochastic volatility. It includes several classical models\, such as the Cox–Ingersoll–Ross model\, as special cases.\n\nIn this talk\, we discuss the estimation of drift parameters in CKLS diffusion models based on continuous-time observations. We focus on the construction of explicit estimators using the method of moments and investigate their asymptotic p roperties. In particular\, we address strong consistency\, asymptotic norm ality\, existence of moments\, and convergence in L^p under suitable assum ptions. We also discuss the role of the model parameters in the asymptotic behavior of the estimators and illustrate the theoretical results with nu merical simulations.\n LOCATION:https://researchseminars.org/talk/AMIS/17/ URL:https://knu-ua.zoom.us/j/89643295643?pwd=eTBZZSt0d0thZzFyaUhDUFNGTVE3Q T09 Passcode (if necessary) 785163 END:VEVENT BEGIN:VEVENT SUMMARY:PhD student Daniel Horbunov (Taras Shevchenko National University of Kyiv and EPFL\, Lausanne\, Switzerland) DTSTART:20261028T150000Z DTEND:20261028T160000Z DTSTAMP:20261025T092952Z UID:AMIS/18 DESCRIPTION:Title: Ne arest neighbour bandwidth choice for nonparametric regression on mixtures with varying concentrations\nby PhD student Daniel Horbunov (Taras Sh evchenko National University of Kyiv and EPFL\, Lausanne\, Switzerland) as part of Asymptotic Methods in Statistics\n\nInteractive livestream: https ://knu-ua.zoom.us/j/89643295643?pwd=eTBZZSt0d0thZzFyaUhDUFNGTVE3QT09 Pass code (if necessary) 785163\n\nAbstract\nFinite mixture models naturally ar ise in statistical analysis of biological and sociological data. If the su bpopulation which a subject belongs to is not known exactly\, the distribu tion of its variables is a mixture of the subpopulations’ distributions. In the classical finite mixture models (FMM) the concentrations of the co mponents in the mixture (mixing probabilities) are the same for all observ ations. In a more flexible mixture with varying concentrations model (MVC) \, the concentrations are different for different observations.\n\nRegress ion models are typically applied to describe dependency between different numerical variables of one subject. In the case of homogeneous sample ther e exist many nonparametric estimators of the regression function\, such as the Nadaraya-Watson estimator (NWE) and local linear regression estimator (LLRE). For homogeneous samples\, NWE demonstrates an inappropriate bias in points where the regressor probability density function (PDF) has disco ntinuity (jump points). For such scenario\, the LLRE stands as a remedy\, having a significantly smaller bias.\n\nIn case of the MVC\, we consider a modification of NWE (mNWE) and LLRE (mLLRE) for estimation of the regress ion function of a given component. Under suitable assumptions\, these esti mators are asymptotically normal\, which makes it possible to derive theor etically optimal bandwidths. However\, such bandwidths depend on unknown p arameters of the underlying distributions and therefore cannot be used in practice.\n\nIn this talk\, we propose an empirical bandwidth choice for t he mNWE and mLLRE based on a modification of the nearest neighbour techniq ue. A simulation study shows that this approach performs satisfactorily co mpared with the theoretically optimal bandwidth\, as measured by the asymp totic mean squared error (aMSE) or the asymptotic mean integrated squared error (aMISE).\n LOCATION:https://researchseminars.org/talk/AMIS/18/ URL:https://knu-ua.zoom.us/j/89643295643?pwd=eTBZZSt0d0thZzFyaUhDUFNGTVE3Q T09 Passcode (if necessary) 785163 END:VEVENT END:VCALENDAR