Testing the Equality of Covariance Operators in Functional Samples

Abstract. We propose a non‐parametric test for the equality of the covariance structures in two functional samples. The test statistic has a chi‐square asymptotic distribution with a known number of degrees of freedom, which depends on the level of dimension reduction needed to represent the data. Detailed analysis of the asymptotic properties is developed. Finite sample perfo‐rmance is examined by a simulation study and an application to egg‐laying curves of fruit flies.     

Distribution of Increases in Residual Log Likelihood for Nested Spatial Models

Here we review nested relationships between models in the Matérn family of spatial models. The problem of comparing nested statistical models is straightforward in regular parametric problems via the likelihood ratio statistics and its asymptotic distribution. Here we examine the distribution of increments in residual log likelihood between nested spatial models when the null hypothesis is that the spatial structure is a convex combination of white noise and the de Wijs process, also known by its logarithmic covariance function. This study is carried out by simulation of spatial processes and the important aspects of this work include how to simulate a spatial process of order 0, the lack of strong bias in the estimates of variance components, and the validity of the usual asymptotic results for nested spatial models examined here.     

Necessary and Sufficient Condition for Nonsingular Fisher Information Matrix in ARMA and Fractional ARIMA Models

It is demonstrated that a necessary and sufficient condition for the Fisher information matrix of a causal and invertible ARMA to be nonsingular is that the model not be redundant; that is, the autoregressive and moving-average polynomials have no roots in common. This result is also extended to fractional ARIMA models.     

A Comparison of Utilized and Theoretical Covariance Weighting Matrices on the Estimation Performance of Quadratic Inference Functions

The quadratic inference function (QIF) method is increasingly popular for the marginal analysis of correlated data due to its advantages over generalized estimating equations. Asymptotic theory is used to derive analytical results from the QIF, and we, therefore, study three asymptotically equivalent weighting matrices in terms of finite-sample parameter estimation. Furthermore, to improve small-sample estimation, we study modifications to the estimation procedure. Examples are presented via simulations and application. Results show that although theoretical weighting matrices work best, the proposed estimation procedure, in which initial estimates are held constant within the matrix of estimated empirical covariances, is preferable in practice.     

The johnson translation system in monte carlo studies

Multivariate lognormal and sinh^-1-normal random vectors, members of the Johnson family, can be generated by direct transformation of multinormal random vectors. Formulae for specifying the parameters of the multinomial distribution in terms of the first and second order moments of the desired multivariate lognormal or sinh^-1-normal distribution are given. These results facilitate the use of these distributions in Monte Carlo studies. In some previous studies employing these distributions, the effect of non-normality was confounded with that of unequal covariance structure.     

Repeated measurements,log likelihood ratio tests and small samples

Various models have previously been proposed for data comprising m repeated measurements on each of N subjects. Log likelihood ratio tests may be used to help choose between possible models, but these tests are based on distributions which in theory apply only asymptotically. With small N , the log likelihood ratio approximation is unreliable, tending to reject the simpler of two models more often than it should. This is shown by reference to three datasets and analogous simulated data. For two of the three datasets, subjects fall into two groups. Log likelihood ratio tests confirm that for each of these two datasets group means over time differ. Tests suggest that group covariance structures also differ.     

Skew-Normal ARMA Models with Nonlinear Heteroscedastic Predictors

Multivariate skew-normal (SN) distributions (Azzalini and Dalla Valle, 1996 Azzalini , A. , Dalla Valle , A. ( 1996 ). The multivariate skew-normal distribution . Biometrika 83 : 715 – 726 .[Crossref], [Web of Science ®] , [Google Scholar]) enjoy some of the useful properties of normal distributions, have nonlinear heteroscedastic predictors but lack the closure property of normal distributions (the sum of independent SN random variables is not SN). Recently, there has been a proliferation of classes of SN distributions with certain closure properties, one of the most promising being the closed skew-normal (CSN) distributions of González-Farías et al. (2004 González-Farías , G. , Dominguez-Molina , J. A. , Gupta , A. K. ( 2004 ). Additive properties of skew-normal random vectors . J. Statist. Plann. Infer. 126 : 521 – 534 .[Crossref], [Web of Science ®] , [Google Scholar]). We study the construction of stationary SN ARMA models for colored SN noise and show that their finite-dimensional distributions are skew-normal, seldom strictly stationary and their covariance functions differ from their normal ARMA counterparts in that they do not converge to zero for large lags. The situation is better for ARMA models driven by CSN noise, but at the additional cost of considerable computational complexity and a less explicit skewness parameter. In view of these results, the widespread use of such classes of SN distributions in the framework of ARMA models seem doubtful.     

Estimation of Variance Components for a Linear Toeplitz Model

The linear Toeplitz covariance structure model of order one is considered. We give some elegant explicit expressions of the Locally Minimum Variance Quadratic Unbiased Estimators of its covariance parameters. We deduce from a Monte Carlo method some properties of their Gaussian maximum likelihood estimators. Finally, for small sample sizes, these two types of estimators are compared with the intuitive empirical estimators and it is shown that the empirical biased estimators should be used.     

Limiting Spectral Distribution for Large Sample Covariance Matrices with m-Dependent Elements

This article establishes the limiting spectral distribution of large sample covariance matrices with m-dependent random variables under the second moment condition by verifying the condition of Theorem 1.1 in Bai and Zhou (2008 Bai , Z. D. , Zhou , W. , ( 2008 ). Large sample covariance matrices without independence structure in columns . Statist. Sinica 2 : 425 – 443 . [Google Scholar]).     

Varying-Coefficient Functional Linear Regression Models

This article considers a generalization of the functional linear regression in which an additional real variable influences smoothly the functional coefficient. We thus define a varying-coefficient regression model for functional data. We propose two estimators based, respectively, on conditional functional principal regression and on local penalized regression splines and prove their pointwise consistency. We check, with the prediction one day ahead of ozone concentration in the city of Toulouse, the ability of such nonlinear functional approaches to produce competitive estimations.