Covariance function | EPFL Graph Search

In probability theory and statistics, the covariance function describes how much two random variables change together (their covariance) with varying spatial or temporal separation. For a random field or stochastic process Z(x) on a domain D, a covariance function C(x, y) gives the covariance of the values of the random field at the two locations x and y: :C(x,y):=\operatorname{cov}(Z(x),Z(y))=\mathbb{E}\left[{Z(x)-\mathbb{E}[Z(x)]}\cdot{Z(y)-\mathbb{E}[Z(y)]} \right]., The same C(x, y) is called the autocovariance function in two instances: in time series (to denote exactly the same concept except that x and y refer to locations in time rather than in space), and in multivariate random fields (to refer to the covariance of a variable with itself, as opposed to the cross covariance between two different variables at different locations, Cov(Z(x1), Y(x2))). Admissibility For locations x1, x2, …, xN ∈ D the variance of every line

Functional data analysis by matrix completion

Marie-Hélène Descary

Traditional approaches to analysing functional data typically follow a two-step procedure, consisting in first smoothing and then carrying out a functional principal component analysis. The idea underlying this procedure is that functional data are well approximated by smooth functions, and that rough variations are due to noise. However, it may very well happen that localised features are rough at a global scale but still smooth at some finer scale. In this thesis we put forward a new statistical approach for functional data arising as the sum of two uncorrelated components: one smooth plus one rough. We give non-parametric conditions under which the covariance operators of the smooth and of the rough components are jointly identifiable on the basis of discretely observed data: the covariance operator corresponding to the smooth component must be of finite rank and have real analytic eigenfunctions, while the one corresponding to the rough component must have a banded covariance function. We construct consistent estimators of both covariance operators without assuming knowledge of the true rank or bandwidth. We then use them to estimate the best linear predictors of the the smooth and the rough components of each functional datum. In both the identifiability and the inference part, we do not follow the usual strategy used in functional data analysis which is to first employ smoothing and work with continuous estimate of the covariance operator. Instead, we work directly with the covariance matrix of the discretely observed data, which allows us to use results and tools from linear algebra. In fact, we show that the whole problem of uniquely recovering the covariance operator of the smooth component given the one of the raw data can be seen as a low-rank matrix completion problem, and we make great use of a classical relation between the rank and the minors of a matrix to solve this matrix completion problem. The finite-sample performance of our approach is studied by means of simulation study.

EPFL2017

Recovering covariance from functional fragments

Marie-Hélène Descary, Victor Panaretos

We consider nonparametric estimation of a covariance function on the unit square, given a sample of discretely observed fragments of functional data. When each sample path is observed only on a subinterval of length , one has no statistical information on the unknown covariance outside a -band around the diagonal. The problem seems unidentifiable without parametric assumptions, but we show that nonparametric estimation is feasible under suitable smoothness and rank conditions on the unknown covariance. This remains true even when the observations are discrete, and we give precise deterministic conditions on how fine the observation grid needs to be relative to the rank and fragment length for identifiability to hold true. We show that our conditions translate the estimation problem to a low-rank matrix completion problem, construct a nonparametric estimator in this vein, and study its asymptotic properties. We illustrate the numerical performance of our method on real and simulated data.

2019

Decompositions Of Log-Correlated Fields With Applications

Janne Matias Junnila

In this article, we establish novel decompositions of Gaussian fields taking values in suitable spaces of generalized functions, and then use these decompositions to prove results about Gaussian multiplicative chaos. We prove two decomposition theorems. The first one is a global one and says that if the difference between the covariance kernels of two Gaussian fields, taking values in some Sobolev space, has suitable Sobolev regularity, then these fields differ by a Holder continuous Gaussian process. Our second decomposition theorem is more specialized and is in the setting of Gaussian fields whose covariance kernel has a logarithmic singularity on the diagonal-or log-correlated Gaussian fields. The theorem states that any log-correlated Gaussian field X can be decomposed locally into a sum of a Holder continuous function and an independent almost *-scale invariant field (a special class of stationary log-correlated fields with 'cone-like' white noise representations). This decomposition holds whenever the term g in the covariance kernel C-X(x, y) = log(1/vertical bar x - y vertical bar) + g(x, y) has locally Hd+epsilon Sobolev smoothness. We use these decompositions to extend several results that have been known basically only for *-scale invariant fields to general log-correlated fields. These include the existence of critical multiplicative chaos, analytic continuation of the subcritical chaos in the so-called inverse temperature parameter beta, as well as generalised Onsager-type covariance inequalities which play a role in the study of imaginary multiplicative chaos.

2019

Functional data analysis by matrix completion

Marie-Hélène Descary

EPFL2017