Tests for separability in nonparametric covariance operators of random surfaces

The assumption of separability of the covariance operator for a random image or hypersurface can be of substantial use in applications, especially in situations where the accurate estimation of the full covariance structure is unfeasible, either for computational reasons, or due to a small sample size. However, inferential tools to verify this assumption are somewhat lacking in high-dimensional or functional {data analysis} settings, where this assumption is most relevant. We propose here to test separability by focusing on $K$-dimensional projections of the difference between the covariance operator and a nonparametric separable approximation. The subspace we project onto is one generated by the eigenfunctions of the covariance operator estimated under the separability hypothesis, negating the need to ever estimate the full non-separable covariance. We show that the rescaled difference of the sample covariance operator with its separable approximation is asymptotically Gaussian. As a by-product of this result, we derive asymptotically pivotal tests under Gaussian assumptions, and propose bootstrap methods for approximating the distribution of the test statistics. We probe the finite sample performance through simulations studies, and present an application to log-spectrogram images from a phonetic linguistics dataset.

[1]  P. Guttorp,et al.  Geostatistical Space-Time Models, Stationarity, Separability, and Full Symmetry , 2007 .

[2]  Jane-Ling Wang,et al.  Review of Functional Data Analysis , 2015, 1507.05135.

[3]  Sean L Simpson,et al.  Separability tests for high-dimensional, low-sample size multivariate repeated measures data , 2014, Journal of applied statistics.

[4]  H. Müller,et al.  Modelling function‐valued stochastic processes, with applications to fertility dynamics , 2017 .

[5]  Piotr Kokoszka P. Secchi, S. Vantini and V. Vitelli: Analysis of spatio-temporal mobile phone data: a case study in the metropolitan area of Milan , 2015, Stat. Methods Appl..

[6]  Simone Vantini,et al.  Analysis of spatio-temporal mobile phone data: a case study in the metropolitan area of Milan , 2015, Statistical Methods & Applications.

[7]  James O. Ramsay,et al.  Applied Functional Data Analysis: Methods and Case Studies , 2002 .

[8]  R. Ryan Introduction to Tensor Products of Banach Spaces , 2002 .

[9]  Marc G. Genton,et al.  Testing for separability of space–time covariances , 2005 .

[10]  J. Aston,et al.  Evaluating stationarity via change-point alternatives with applications to fMRI data , 2012, 1301.2894.

[11]  Surajit Ray,et al.  Functional principal component analysis of spatially correlated data , 2014, Stat. Comput..

[12]  D. Zimmerman,et al.  The likelihood ratio test for a separable covariance matrix , 2005 .

[13]  M. Lindquist The Statistical Analysis of fMRI Data. , 2008, 0906.3662.

[14]  Karl J. Friston,et al.  A unified statistical approach for determining significant signals in images of cerebral activation , 1996, Human brain mapping.

[15]  H. Müller,et al.  Pairwise curve synchronization for functional data , 2008 .

[16]  Robert Tibshirani,et al.  In praise of sparsity and convexity , 2013 .

[17]  Hans-Georg Müller,et al.  Functional Data Analysis , 2016 .

[18]  Piercesare Secchi,et al.  Distances and inference for covariance operators , 2014 .

[19]  Sean L. Simpson An Adjusted Likelihood Ratio Test for Separability in Unbalanced Multivariate Repeated Measures Data , 2010 .

[20]  P. Kokoszka,et al.  Testing separability of space--time functional processes , 2015, 1509.07017.

[21]  P. Hall,et al.  On properties of functional principal components analysis , 2006 .

[22]  I. Gohberg,et al.  Classes of Linear Operators , 1990 .

[23]  Damien Garcia,et al.  Robust smoothing of gridded data in one and higher dimensions with missing values , 2010, Comput. Stat. Data Anal..

[24]  R. Kadison,et al.  Fundamentals of the Theory of Operator Algebras , 1983 .

[25]  J. Ringrose Compact non-self-adjoint operators , 1971 .

[26]  M. Genton Separable approximations of space‐time covariance matrices , 2007 .

[27]  I. Gohberg Classes of Linear Operators, Vol.I, Operator Theory , 1990 .

[28]  Z. Q. John Lu,et al.  Nonparametric Functional Data Analysis: Theory And Practice , 2007, Technometrics.

[29]  Marie Frei,et al.  Functional Data Analysis With R And Matlab , 2016 .

[30]  Frédéric Ferraty,et al.  Nonparametric Functional Data Analysis: Theory and Practice (Springer Series in Statistics) , 2006 .

[31]  Past, Present, and Future of Statistical Science , 2015 .

[32]  Elliott Ward Cheney,et al.  Multivariate approximation theory , 1986 .

[33]  E. Cheney Multivariate approximation theory: selection topics , 1986 .

[34]  Ronald W. Schafer,et al.  Digital Processing of Speech Signals , 1978 .

[35]  Henry W. Altland,et al.  Applied Functional Data Analysis , 2003, Technometrics.

[36]  N. Cressie,et al.  Classes of nonseparable, spatio-temporal stationary covariance functions , 1999 .

[37]  Piotr Kokoszka,et al.  Inference for Functional Data with Applications , 2012 .

[38]  J. O. Ramsay,et al.  Functional Data Analysis (Springer Series in Statistics) , 1997 .

[39]  M. Fuentes Testing for separability of spatial–temporal covariance functions , 2006 .

[40]  H. Müller,et al.  Functional Data Analysis for Sparse Longitudinal Data , 2005 .

[41]  Luo Xiao,et al.  Two-way principal component analysis for matrix-variate data, with an application to functional magnetic resonance imaging data. , 2016, Biostatistics.

[42]  André Mas,et al.  A sufficient condition for the CLT in the space of nuclear operators--Application to covariance of random functions , 2006 .

[43]  T. Gneiting Nonseparable, Stationary Covariance Functions for Space–Time Data , 2002 .