Time series and statistical learning

Non-parametric statistics, causal inference, high-dimensional statistics

Mona Azadkia's research is centered on advancing methodologies to analyse and quantify the dependency structure in data. Her work focuses on developing interpretable measures of dependence and conditional dependence between variables, which play a key role in various statistical tasks such as variable selection, dimensionality reduction, sensitivity analysis, causal inference, and hypothesis testing.

Change-point, shape constraint, computing, nonparametric, time series

Yining's current research focuses on developing new methods for statistical problems such as change-point detection and nonparametric estimation. He is also interested in understanding the computational aspects of statistical methods. He completed his PhD (2014) in Statistics at the University of Cambridge.

Please note Yining is on sabbatical for all of 2025/26.

Nonparametric and shape-constrained inference, approximation message passing, score matching, survival data analysis

Oliver is a Assistant Professor in Statistics, having previously been a lecturer at the University of Bath and a PhD student and postdoc at the University of Cambridge. He is interested in developing flexible theory and methodology that respect meaningful structural constraints. For instance, he has worked on shape-restricted inference problems where it may be reasonable to assume that a regression function or log-density is concave, but not that it belongs to a specific parametric family. Oliver is also exploring the applications of score matching and its links to convex optimisation. Overall, his research has a nonparametric flavour, with a particular emphasis on the robustness of statistical procedures to outliers and model misspecification. Oliver has also been involved in some interdisciplinary collaborations related to respiratory illnesses and the career intentions of medical students.

Time series, change-points, multiscale methods, causality, machine learning

Piotr Fryzlewicz's research interests lie in multiscale modelling and estimation, time series (especially nonstationary time series), change-point detection, high-dimensional statistical inference and dimension reduction, statistical learning, networks, functional programming in data science, statistics in finance, statistics in the social sciences, and statistics in neuroscience. Amongst others, Piotr is the originator of the Haar-Fisz transformation and smoothing methodology for non-Gaussian data and the Wild Binary Segmentation method for multiple change-point detection, as well as having co-authored work on Sparsified Binary Segmentation for high-dimensional time series segmentation and tilted correlation for variable selection in high-dimensional regression models.

In 2013, Piotr was awarded a Guy Medal in Bronze by the Royal Statistical Society. Prior to the LSE, Piotr worked at Winton Capital Management, the University of Bristol and Imperial College. He has a PhD in Statistics from the University of Bristol, awarded in 2003.

Bayesian inference, latent stochastic processes, sequential learning, stochastic epidemic modelling, volatility estimation, bond risk premia

Kostas’ research focuses on developing and applying advanced computational methods, such as Markov Chain and Sequential Monte Carlo, for Bayesian Inference. His methodology has mostly targeted continuous time probability models based on stochastic differential equations driven by standard or fractional Brownian motion. The areas of application include Financial and Econometric Time Series as well biomedical problems such as stochastic epidemic models and analysis of growth curves.

Prior to joining the Statistics Department of LSE, he was a post-doctoral researcher at the University of Cambridge, in the Signal Processing Laboratory of the Engineering Department. He completed his PhD (2007) in the Statistics Department of the Athens University of Economics and Business while spending some time at University of Lancaster.

Multiple testing, change-point detection, model selection techniques, sparse modelling and nonparametric estimation

Anica's main research interests lie in multiple testing, change-point detection, model selection techniques, sparse modelling and nonparametric estimation. She is received her PhD from LSE in 2022.

Financial time series, spatial econometrics, tensor time series, dimension reduction, factor modelling

Clifford’s research are focused in 1) statistical leaning techniques, especially for high dimensional data, and 2) time series analysis. These include semiparametric modelling, variables and feature selection, regularization methods for high dimensional time series analysis, to name but a few areas. One particular area of interest is the estimation of a large covariance/precision matrix from data. With random matrix theories more developed over the past decade, it is a high time for further developments of theories and methodologies in the area of high dimensional matrix estimation and applications. This area is important in a wide variety of scientific fields, including portfolio allocation and risk assessment in finance, classification and large scale hypothesis testing in bioinformatics, forecasting in macroeconomics time series, or cosmological survey in astrophysics.

High dimensional time series modelling is needed nowadays for making sense of vast volume of data we can find everyday around us, including the internet. Clifford Lam’s one particular research interest about high dimensional time series analysis is to find "factors" that can explain/summarise a majority of time series dynamics, like market factors that explain the dynamics of most stock prices in FTSE 100. A goal is to incorporate large number of variables that we can observe, and then summarise these into factors for increasing the explanatory and predictive power of various time series models.

Another area of research is in spatial econometrics modelling. A commonly used model is the spatial lag/error model for a large spatial panel of time series. To be able to use such a model, a component called the spatial weight matrix, which is a square matrix of the same size as the dimension of the panel, has to be pre-specified. This matrix represents the underlying spatial interdependence structure of the data. Yet there are no universal rules in doing so, and most practitioner use certain distance metrics to specify such a matrix, which is at best a crude approximation to the said structure. Moreover, when the dimension of the panel increases, it becomes more difficult to specify this spatial weight matrix. There are in fact various ways to estimate the said matrix from data, and inference in the models with such a matrix estimated rather than pre-specified is also an important research element.

Prior to joining the LSE in 2008, Clifford was a PhD student in the department of Operations Research & Financial Engineering, Princeton University.

Functional data analysis, functional time series, high-dimensional time series, high-dimensional statistical inference, large-scale multiple testing

Xinghao’s research is focused on (i) functional and longitudinal data analysis, (ii) high dimensional statistical inference, e.g. covariance and precision matrix estimation, variable selection, (iii) time series analysis, e.g. functional time series, high dimensional time series, (iv) statistical machine learning with applications in Business, Neuroimaging Analysis and Environmental Sciences.

Prior to joining the LSE as an Assistant Professor in Statistics, Xinghao earned his PhD in Business Statistics from Marshall School of Business at the University of Southern California, M.S. in Statistics at the University of Chicago and B.S. in Mathematics and Physics at Tsinghua University.

High-dimensional data, change point analysis, sparse signal detection, robust statistics, dimension reduction

Tengyao Wang is broadly interested in the area of high-dimensional statistics. His research focuses mainly on developing computationally efficient procedures for high-dimensional problems, while at the same time understanding the potential statistical limitations imposed by computational constraints. Some of his current research interests include: (i) Sparse signal detection in high-dimensional data; (ii) Change-point detection and estimation problems; (iii) Dimension reduction techniques; (iv) Robust statistical procedures in the presence of missing data or heavy-tailed noise; (v) Nonparametric statistical inference and (vi) Applications, including medical statistics, financial data analysis and statistical learning-assisted material discovery.

Prior to joining LSE as an associate professor, Tengyao was a lecturer at University College London and a research fellow at Cantab Capital Institute for the Mathematics of Information, University of Cambridge. Tengyao was awarded the Royal Statistical Society Research Prize in 2019 and the Guy Medal in Bronze in 2023.

High-dimensional time series analysis, dimension reduction and factor modelling, dynamic network, spatio-temporal processes, nonstationary processes and cointegration

Professor Yao's recent research has mainly (but not exclusively) focused on analysing complex and high-dimensional time series in the sense that the observation recorded at each time is more complex than a single scalar or a short vector. Two examples under this framework are spatio-temporal data and dynamic network for which the observation at each time is a space or a network. Demand for analysing those data on an ever-increasing scale is a part of the real challenge underneath the buzzword BigData. The time series thinking and methodology can contribute in facing those challenges. An overarching aim has been to develop new tools which reduce the dimension and/or the complexity of time series by exploring latent low dimensional structures.

He has also published in the areas of nonlinear time series, functional time series, econometrics, extreme values of dependent data, non/semi-parametric estimation for conditional distributions, non/semi-parametric regression. Take a look at a full list of publications.

Professor Yao also enjoys collaborating with industry, which often leads to new, non-standard and challenging research problems.

With the increasing experience in research and, especially, in solving practical problems arising from consultancy, Professor Yao is more attracted to the pursuing for the simplicity in statistical methodology, as simple methods are often effective, and gain more appreciation in application. With the ever-increasing data size and complexity in this information age, the challenge is to develop simple, if ever possible, statistical methods relevant to solving new and complex practical problems.

Please note Qiwei is on sabbatical for all of 2025/26.