arXiv subjects

Marina I. Knight

Publications and source records attributed to Marina I. Knight.

14 recordsLinked to original sources

arXiv2026

Multiscale Dynamic Dependence Estimation over Networks

In many settings, observed multivariate time series are often nonstationary in nature, i.e., their second order properties vary over time. An additional feature is that their cross-channel dependencies are structured by an underlying network. Together, they give rise to complex interactions between temporal dynamics and network topology. We propose Locally Stationary Wavelet processes on Networks (Net-LSW), a new framework for modelling multiscale, time-varying dependencies that explicitly incorporates the network structure. Unlike traditional multivariate approaches, the Net-LSW process encodes the graph directly in the covariance structure of its random increments. We introduce the concept of local partial correlation graph, mathematically connecting absent edges to zero entries in the time-scale dependent inverse wavelet spectral structure. For inference on the local cross-nodal (partial) dependence, we develop a novel subprocess-based estimation scheme and establish its consistency properties. This new pipeline for network-based nonstationary process modelling, complete with estimation and simulation capabilities that extend outside time-varying vector autoregressive models, is shown to accurately recover evolving dependence structures whilst respecting the underlying graph topology. The analysis of daily stock price volatilities across a global bank network captures multiscale, highly nonstationary dependencies and identifies time-varying systemic shifts during major financial shocks, including Brexit and the COVID-19 pandemic.

stat.ME↗

arXiv2026

Partial Wavelet Canonical Coherence for Nonstationary Signals with High Dimensional Confounders

We develop Partial Wavelet Canonical Coherence for measuring the direct canonical association between two multivariate nonstationary time series after adjustment for possibly high-dimensional confounders. To the best of our knowledge, this is the first method that establishes a frequency-domain formulation of the partial canonical correlation analysis for time series. Through a wavelet approach, the proposed method yields a scale-specific, time-varying measure of association capable to work with potential data nonstationarities. We formulate the target quantity under the multivariate locally stationary wavelet framework, develop principled estimation through local wavelet spectral matrices, and incorporate principal-component reduction for stable adjustment in high-dimensions. Simulations show that the method removes spurious marginal association induced by confounding and accurately recovers direct association, including in higher-dimensional settings. Analysis of U.S. exchange-traded funds reveals substantial time-varying and scale-dependent direct canonical association after adjustment for external market effects.

stat.ME↗

arXiv2026

Dynamic cross-scale wavelet coherence

This paper develops a novel statistical approach that allows for the {\em first time} the {\em cross}-oscillatory characterisation of temporally localised interactions between channels in a functional brain network. Brain signals are often nonstationary and the proposed framework uses wavelets as an effective tool for capturing (i) single-scale channel transient features, due to their adaptiveness to the dynamic signal properties, and (ii) cross-scale channel interactions, due to their multiscale nature. Our approach introduces scale-specific {\em subprocesses} and {\em cross-scale (CS) dependencies} for a new class of multivariate locally stationary (MvLSW) wavelet processes that we refer to as CS-MvLSW. Under this new model, we develop two consistent estimation procedures for the {\em localised} single- and cross-scale channel dependence. Extensive simulation studies demonstrate that the theoretically established properties hold in practice. The proposed CS-MvLSW framework remains accurate under pronounced cross-scale dependence, whereas existing MvLSW coherence estimates dramatically deteriorate even for single-scales when such complex structure is present. The proposed approach was used for electroencephalogram (EEG) data to study alterations in the functional connectivity structure in children diagnosed with attention deficit hyperactivity disorder (ADHD), and identified novel clinically pertinent cross-scale interactions in the functional brain network across the left and right hemispheres, differentiating brain connectivity between control and ADHD groups.

stat.ME↗

arXiv2026

Frequency-Domain Analysis of Time Series with Network-Structured Dependence: Application to Global Bank Connectedness

Financial spillovers in interconnected systems, such as global banking networks, require tools that capture temporal and frequency dynamics, while incorporating the underlying network topology. While current network time series models are developed in the time-domain, frequency-domain approaches, which reveal how cross-nodal dependencies vary across different cycles, remain under-explored. This paper develops a spectral analysis framework that accommodates flexible forms of network dependence, including interactions mediated through intermediate nodes. This ensures that inter-nodal relationships are not restricted to direct connections, a feature crucial for capturing indirect financial spillovers. We define the network time series spectral density, alongside coherence and partial coherence, and propose both parametric and network-constrained nonparametric methods for their estimation. Simulations and theoretical results demonstrate the strong performance of the parametric approach when the data-generating process aligns with the model structure, whereas the nonparametric alternative provides robustness against model misspecification. An application to global bank connectedness shows that the proposed spectral measures capture inter-bank frequency-specific spillover effects, yielding results consistent with existing measures while additionally uncovering richer patterns of volatility transmission that are intimately connected to the network topology.

stat.ME↗

arXiv2025

Network Estimation for Stationary Time Series

High-dimensional multivariate time series are common in many scientific and industrial applications, where the interest lies in identifying key dependence structure within the data for subsequent analysis tasks, such as forecasting. An important avenue to achieve this is through the estimation of the conditional independence graph via graphical models, although for time series data settings the underpinning temporal dependence can make this task challenging. In this article, we propose a novel wavelet domain technique that allows the data-driven inference of the (sparse) conditional independence graph of a high-dimensional stationary multivariate time series. By adopting the locally stationary wavelet modelling framework, we repose the estimation problem as a well-principled wavelet domain graphical lasso formulation. Theoretical results establish that our associated estimation scheme enjoys good consistency properties when determining sparse dependence structure in input time series data. The performance of the proposed method is illustrated using extensive simulations and we demonstrate its applicability on a real-world dataset representing hospitalisations of COVID-19 patients.

stat.ME↗

arXiv2025

Long memory network time series

Many scientific areas, from computer science to the environmental sciences and finance, give rise to multivariate time series which exhibit long memory, or loosely put, a slow decay in their autocorrelation structure. Efficient modelling and estimation in such settings is key for a number of analysis tasks, such as accurate prediction. However, traditional approaches for modelling such data, for example long memory vector autoregressive processes, are challenging even in modest dimensions, as the number of parameters grows quadratically with the number of modelled variables. Additionally, in many practical data settings, the observed series is accompanied by a (possibly inferred) network that provides information about the presence or absence of between-component associations via the graph edge topology. This article proposes two new models for capturing the dynamics of long memory time series where a network is accounted for. Our approach not only facilitates the analysis of graph-structured long memory time series, but also improves computational efficiency over traditional multivariate long memory models by leveraging the inherent low-dimensional parameter space by adapting likelihood-based estimation algorithms to the network setting. Simulation studies show that our proposed estimation is more stable than traditional models, and is able to tackle data scenarios where current models fail due to computational challenges. While widely applicable, here we demonstrate the efficacy of our proposed models on datasets arising in environmental science and finance.

stat.ME↗

arXiv2025

Wavelet Canonical Coherence for Nonstationary Signals

Understanding the evolving dependence between two clusters of multivariate signals is fundamental in neuroscience and other domains where sub-networks in a system interact dynamically over time. Despite the growing interest in multivariate time series analysis, existing methods for between-clusters dependence typically rely on the assumption of stationarity and lack the temporal resolution to capture transient, frequency-specific interactions. To overcome this limitation, we propose scale-specific wavelet canonical coherence (WaveCanCoh), a novel framework that extends canonical coherence analysis to the nonstationary setting by leveraging the multivariate locally stationary wavelet model. The proposed WaveCanCoh enables the estimation of time-varying canonical coherence between clusters, providing interpretable insight into scale-specific time-varying interactions between clusters. Through extensive simulation studies, we demonstrate that WaveCanCoh accurately recovers true coherence structures under both locally stationary and general nonstationary conditions. Application to local field potential (LFP) activity data recorded from the hippocampus reveals distinct dynamic coherence patterns between correct and incorrect memory-guided decisions, illustrating the capacity of the method to detect behaviorally relevant neural coordination. These results highlight WaveCanCoh as a flexible and principled tool for modeling complex cross-group dependencies in nonstationary multivariate systems. The code for WaveCanCoh is available at: https://github.com/mhaibo/WaveCanCoh.git.

stat.ME↗

arXiv2025

Continuous Time Locally Stationary Wavelet Processes

This article introduces the class of continuous time locally stationary wavelet processes. Continuous time models enable us to properly provide scale-based time series models for irregularly-spaced observations for the first time, while also permitting a spectral representation of the process over a continuous range of scales. We derive results for both the theoretical setting, where we assume access to the entire process sample path, and a more practical one, which develops methods for estimating the quantities of interest from sampled time series. The latter estimates are accurately computable in reasonable time by solving the relevant linear integral equation using the iterative soft-thresholding algorithm due to Daubechies, Defrise and De~Mol. Appropriate smoothing techniques are also developed and applied in this new setting. Comparisons to previous methods are conducted on the heart rate time series of a sleeping infant. Additionally, we exemplify our new methods by computing spectral and autocovariance estimates on irregularly-spaced heart rate data obtained from a recent sleep-state study.

math.ST↗

arXiv2024

A multiscale method for data collected from network edges via the line graph

Data collected over networks can be modelled as noisy observations of an unknown function over the nodes of a graph or network structure, fully described by its nodes and their connections, the edges. In this context, function estimation has been proposed in the literature and typically makes use of the network topology such as relative node arrangement, often using given or artificially constructed node Euclidean coordinates. However, networks that arise in fields such as hydrology (for example, river networks) present features that challenge these established modelling setups since the target function may naturally live on edges (e.g., river flow) and/or the node-oriented modelling uses noisy edge data as weights. This work tackles these challenges and develops a novel lifting scheme along with its associated (second) generation wavelets that permit data decomposition across the network edges. The transform, which we refer to under the acronym LG-LOCAAT, makes use of a line graph construction that first maps the data in the line graph domain. We thoroughly investigate the proposed algorithm's properties and illustrate its performance versus existing methodologies. We conclude with an application pertaining to hydrology that involves the denoising of a water quality index over the England river network, backed up by a simulation study for a river flow dataset.

stat.ME↗

arXiv2023

Automatic Locally Stationary Time Series Forecasting with application to predicting U.K. Gross Value Added Time Series under sudden shocks caused by the COVID pandemic

Accurate forecasting of the U.K. gross value added (GVA) is fundamental for measuring the growth of the U.K. economy. A common nonstationarity in GVA data, such as the ABML series, is its increase in variance over time due to inflation. Transformed or inflation-adjusted series can still be challenging for classical stationarity-assuming forecasters. We adopt a different approach that works directly with the GVA series by advancing recent forecasting methods for locally stationary time series. Our approach results in more accurate and reliable forecasts, and continues to work well even when the ABML series becomes highly variable during the COVID pandemic.

stat.ME↗

arXiv2020

Multiscale modelling of replicated nonstationary time series

Within the neurosciences, to observe variability across time in the dynamics of an underlying brain process is neither new nor unexpected. Wavelets are essential in analyzing brain signals because, even within a single trial, brain signals exhibit nonstationary behaviour. However, neurological signals generated within an experiment may also potentially exhibit evolution across trials (replicates). As neurologists consider localised spectra of brain signals to be most informative, here we develop a novel wavelet-based tool capable to formally represent process nonstationarities across both time and replicate dimensions. Specifically, we propose the Replicate Locally Stationary Wavelet (RLSW) process, that captures the potential nonstationary behaviour within and across trials. Estimation using wavelets gives a natural desired time- and replicate-localisation of the process dynamics. We develop the associated spectral estimation framework and establish its asymptotic properties. By means of thorough simulation studies, we demonstrate the theoretical estimator properties hold in practice. A real data investigation into the evolutionary dynamics of the hippocampus and nucleus accumbens during an associative learning experiment, demonstrate the applicability of our proposed methodology, as well as the new insights it provides.

stat.ME↗

arXiv2020

The Local Partial Autocorrelation Function and Some Applications

The classical regular and partial autocorrelation functions are powerful tools for stationary time series modelling and analysis. However, it is increasingly recognized that many time series are not stationary and the use of classical global autocorrelations can give misleading answers. This article introduces two estimators of the local partial autocorrelation function and establishes their asymptotic properties. The article then illustrates the use of these new estimators on both simulated and real time series. The examples clearly demonstrate the strong practical benefits of local estimators for time series that exhibit nonstationarities.

math.ST↗

arXiv2019

Changepoint analysis of historical battle deaths

It has been claimed and disputed that World War II has been followed by a `long peace', an unprecedented decline of war. We conduct a full changepoint analysis of well-documented, publicly-available battle deaths datasets, using new techniques that enable the robust detection of changes in the statistical properties of such heavy-tailed data. We first test and calibrate these techniques. We then demonstrate the existence of changes, independent of data presentation, at around 1910 and 1950 CE, bracketing the World Wars, and around the 1830s and 1994 CE. Our analysis provides a methodology for future investigations and an empirical basis for political and historical discussions.

stat.AP↗

arXiv2016

Clustering nonstationary circadian rhythms using locally stationary wavelet representations

How does soil pollution affect a plant's circadian clock? Are there any differences between how the clock reacts when exposed to different concentrations of elements of the periodic table? If so, can we characterise these differences? We approach these questions by analysing and modelling circadian plant data, where the levels of expression of a luciferase reporter gene were measured at regular intervals over a number of days after exposure to different concentrations of lithium. A key aspect of circadian data analysis is to determine whether a time series (derived from experimental data) is `rhythmic' and, if so, to determine the underlying period. However, our dataset displays nonstationary traits such as changes in amplitude, gradual changes in period and phase-shifts. In this paper, we develop clustering methods using a wavelet transform. Wavelets are chosen as they are ideally suited to identifying discriminant local time and scale features. Furthermore, we propose treating the observed time series as realisations of locally stationary wavelet processes. This allows us to define and estimate the evolutionary wavelet spectrum. We can then compare, in a quantitative way, using a functional principal components analysis, the time-frequency patterns of the time series. Our approach uses a clustering algorithm to group the data according to their time-frequency patterns. We demonstrate the advantages of our methodology over alternative approaches and show that it successfully clusters our data.

stat.AP↗