Search

We consider robust estimation of wrapped models to multivariate circular data that are points on the surface of a $p$-torus based on the weighted likelihood methodology.Robust model fitting is achieved by a set of weighted likelihood estimating equations, based on the computation of data dependent weights aimed to down-weight anomalous values, such as unexpected directions that do not share the main pattern of the bulk of the data. Weighted likelihood estimating equations with weights evaluated on the torus orobtained after unwrapping the data onto the Euclidean space are proposed and compared. Asymptotic properties and robustness features of the estimators under study have been studied, whereas their finite sample behavior has been investigated by Monte Carlo numerical experiment and real data examples.

We propose a MANOVA test for semicontinuous data that is applicable also when the dimensionality exceeds the sample size. The test statistic is obtained as a likelihood ratio, where numerator and denominator are computed at the maxima of penalized likelihood functions under each hypothesis. Closed form solutions for the regularized estimators allow us to avoid computational overheads. We derive the null distribution using a permutation scheme. The power and level of the resulting test are evaluated in a simulation study. We illustrate the new methodology with two original data analyses, one regarding microRNA expression in human blastocyst cultures, and another regarding alien plant species invasion in the island of Socotra (Yemen).

Robust estimators for Generalized Linear Models (GLMs) are not easy to develop because of the nature of the distributions involved. Recently, there has been an increasing interest in this topic, especially in the presence of a possibly large number of explanatory variables. Transformed M-estimators (MT) are a natural way to extend the methodology of M-estimators to the class of GLMs and to obtain robust methods. We introduce a penalized version of MT-estimators in order to deal with high-dimensional data. We prove, under appropriate assumptions, consistency and asymptotic normality of this new class of estimators. The theory is developed for redescending $\rho$-functions and Elastic Net penalization. An iterative re-weighted least squares algorithm is given, together with a procedure to initialize it. The latter is of particular importance, since the estimating equations might have multiple roots. We illustrate the performance of this new method for the Poisson family under several type of contaminations in a Monte Carlo experiment and in an example based on a real dataset., Comment: 48 pages, 4 figures

This paper considers generalised network, intended as networks where (a) the edges connecting the nodes are nonlinear, and (b) stochastic processes are continuously indexed over both vertices and edges. Such topological structures are normally represented through special classes of graphs, termed graphs with Euclidean edges. We build generalised networks in which topology changes over time instants. That is, vertices and edges can disappear at subsequent time instants and edges may change in shape and length. We consider both cases of linear or circular time. For the second case, the generalised network exhibits a periodic structure. Our findings allow to illustrate pros and cons of each setting. Generalised networks become semi-metric spaces whenever equipped with a proper semi-metric. Our approach allows to build proper semi-metrics for the temporally-evolving topological structures of the networks. Our final effort is then devoted to guiding the reader through appropriate choice of classes of functions that allow to build proper reproducing kernels when composed with the temporally-evolving semi-metrics topological structures., Comment: 48 pages, 13 figures

The Seemingly Unrelated Regressions (SUR) model is a wide used estimation procedure in econometrics, insurance and finance, where very often, the regression model contains more than one equation. Unknown parameters, regression coefficients and covariances among the errors terms, are estimated using algorithms based on Generalized Least Squares or Maximum Likelihood, and the method, as a whole, is very sensitive to outliers. To overcome this problem M-estimators and S-estimators are proposed in the literature together with fast algorithms. However, these procedures are only able to cope with row-wise outliers in the error terms, while their performance becomes very poor in the presence of cell-wise outliers and as the number of equations increases. A new robust approach is proposed which is able to perform well under both contamination types as well as it is fast to compute. Illustrations based on Monte Carlo simulations and a real data example are provided., Comment: 18 pages, 6 figures and 3 Tables

A weighted likelihood technique for robust estimation of a multivariate Wrapped Normal distribution for data points scattered on a p-dimensional torus is proposed. The occurrence of outliers in the sample at hand can badly compromise inference for standard techniques such as maximum likelihood method. Therefore, there is the need to handle such model inadequacies in the fitting process by a robust technique and an effective down-weighting of observations not following the assumed model. Furthermore, the employ of a robust method could help in situations of hidden and unexpected substructures in the data. Here, it is suggested to build a set of data-dependent weights based on the Pearson residuals and solve the corresponding weighted likelihood estimating equations. In particular, robust estimation is carried out by using a Classification EM algorithm whose M-step is enhanced by the computation of weights based on current parameters' values. The finite sample behavior of the proposed method has been investigated by a Monte Carlo numerical studies and real data examples., Comment: 18 pages, 4 figures. METRON (2021)

Many real-life data sets can be analyzed using Linear Mixed Models (LMMs). Since these are ordinarily based on normality assumptions, under small deviations from the model the inference can be highly unstable when the associated parameters are estimated by classical methods. On the other hand, the density power divergence (DPD) family, which measures the discrepancy between two probability density functions, has been successfully used to build robust estimators with high stability associated with minimal loss in efficiency. Here, we develop the minimum DPD estimator (MDPDE) for independent but non identically distributed observations in LMMs. We prove the theoretical properties, including consistency and asymptotic normality. The influence function and sensitivity measures are studied to explore the robustness properties. As a data based choice of the MDPDE tuning parameter $\alpha$ is very important, we propose two candidates as "optimal" choices, where optimality is in the sense of choosing the strongest downweighting that is necessary for the particular data set. We conduct a simulation study comparing the proposed MDPDE, for different values of $\alpha$, with the S-estimators, M-estimators and the classical maximum likelihood estimator, considering different levels of contamination. Finally, we illustrate the performance of our proposal on a real-data example., Comment: 25 pages, 4 figures, 3 tables

Local general depth ($LGD$) functions are used for describing the local geometric features and mode(s) in multivariate distributions. In this paper, we undertake a rigorous systematic study of $LGD$ and establish several analytical and statistical properties. First, we show that, when the underlying probability distribution is absolutely continuous with density $f(\cdot)$, the scaled version of $LGD$ (referred to as $\tau$-approximation) converges, uniformly and in $L^d(\mathbb{R}^p)$ to $f(\cdot)$ when $\tau$ converges to zero. Second, we establish that, as the sample size diverges to infinity the centered and scaled sample $LGD$ converge in distribution to a centered Gaussian process uniformly in the space of bounded functions on $\mathcal{H}_G$, a class of functions yielding $LGD$. Third, using the sample version of the $\tau$-approximation ($S \tau A$) and the gradient system analysis, we develop a new clustering algorithm. The validity of this algorithm requires several results concerning the uniform finite difference approximation of the gradient system associated with $S \tau A$. For this reason, we establish \emph{Bernstein}-type inequality for deviations between the centered and scaled sample $LGD$, which is also of independent interest. Finally, invoking the above results, we establish consistency of the clustering algorithm. Applications of the proposed methods to mode estimation and upper level set estimation are also provided. Finite sample performance of the methodology are evaluated using numerical experiments and data analysis., Comment: 39+238 pages, 1+126 figures, 1+15 tables

One of the most common problems that any technique encounters is the high dimensionality of the input data. This yields several problems in the subsequently statistical methods due to so called "curse of dimensionality". Several dimension reduction methods have been proposed in the literature, until now, to accomplish the goal of having smaller dimension space. The most popular among them, is the so called Principal Component Analysis (PCA). One of the extensions of PCA is Probabilistic PCA (known as PPCA). In PPCA, a probabilistic model is used to perform the dimension reduction. By convention, there are cases in applied sciences, e.g. Bioinformatics, Biology and Geology that the data at hand are in non Euclidean space. Elaborating further, an important situation is when each variable poses a direction so that a data point is lying on a torus. According to its importance, the extension of the PCA to such data is being under attention during last decades. In this paper, we introduce a Probabilistic PCA for data on torus (TPPCA), assuming a Multivariate Wrapped Normal distribution as the underline model. Furthermore, we present certain appropriate algorithms to compute this dimension reduction technique and then, we compare our proposal with other dimension reduction techniques using examples based on real datasets and a small Monte Carlo simulation.

In the classical contamination models, such as the gross-error (Huber and Tukey contamination model or Case-wise Contamination), observations are considered as the units to be identified as outliers or not. This model is very useful when the number of considered variables is moderately small. Alqallaf et al. [2009] shows the limits of this approach for a larger number of variables and introduced the Independent contamination model (Cell-wise Contamination) where now the cells are the units to be identified as outliers or not. One approach to deal, at the same time, with both type of contamination is filter out the contaminated cells from the data set and then apply a robust procedure able to handle case-wise outliers and missing values. Here we develop a general framework to build filters in any dimension based on statistical data depth functions. We show that previous approaches, e.g. Agostinelli et al. [2015a] and Leung et al. [2017], are special cases. We illustrate our method by using the half-space depth., Comment: 25 pages, 11 figures. TEST (2021)

The simplicial depth, like other relevant multivariate statistical data depth functions, vanishes right outside the convex hull of the support of the distribution with respect to which the depth is computed. This is problematic when it is required to differentiate among points outside the convex hull of the distribution support, with respect to which the depth is computed, based on their depth values. We provide the first proposal for simplicial depth which do not vanish right outside the convex hull of the distribution. The properties of the proposal and of the corresponding estimator are studied theoretically and by means of Monte Carlo simulations and analysis of datasets., Comment: 26+14 pages, 7+7 figures, 1+2 tables

Centrality descriptors are widely used to rank nodes according to specific concept(s) of importance. Despite the large number of centrality measures available nowadays, it is still poorly understood how to identify the node which can be considered as the `centre' of a complex network. In fact, this problem corresponds to finding the median of a complex network. The median is a non-parametric and robust estimator of the location parameter of a probability distribution. In this work, we present the most natural generalisation of the concept of median to the realm of complex networks, discussing its advantages for defining the centre of the system and percentiles around that centre. To this aim, we introduce a new statistical data depth and we apply it to networks embedded in a geometric space induced by different metrics. The application of our framework to empirical networks allows us to identify median nodes which are socially or biologically relevant.

Multivariate circular observations, i.e. points on a torus are nowadays very common. Multivariate wrapped models are often appropriate to describe data points scattered on p-dimensional torus. However, statistical inference based on this model is quite complicated since each contribution in the log likelihood involve an infinite sum of indices in Z^p where p is the dimension of the problem. To overcome this, two estimates procedures based on Expectation Maximization and Classification Expectation Maximization algorithms are proposed that worked well in moderate dimension size. The performance of the introduced methods are studied by Monte Carlo simulation and illustrated on three real data sets., Comment: 30 pages, 12 figures

A weighted likelihood approach for robust fitting of a mixture of multivariate Gaussian components is developed in this work. Two approaches have been proposed that are driven by a suitable modification of the standard EM and CEM algorithms, respectively. In both techniques, the M-step is enhanced by the computation of weights aimed at downweighting outliers. The weights are based on Pearson residuals stemming from robust Mahalanobis-type distances. Formal rules for robust clustering and outlier detection can be also defined based on the fitted mixture model. The behavior of the proposed methodologies has been investigated by some numerical studies and real data examples in terms of both fitting and classification accuracy and outlier detection.

We propose a general approach to construct weighted likelihood estimating equations with the aim of obtain robust estimates. The weight, attached to each score contribution, is evaluated by comparing the statistical data depth at the model with that of the sample in a given point. Observations are considered regular when the ratio of these two depths is close to one, whereas, when the ratio is large the corresponding score contribution may be downweigthed. Details and examples are provided for the robust estimation of the parameters in the multivariate normal model. Because of the form of the weights, we expect that, there will be no downweighting under the true model leading to highly efficient estimators. Robustness is illustrated using two real data sets.

Convolutional Neural Networks (CNNs) are a popular deep learning architecture widely applied in different domains, in particular in classifying over images, for which the concept of convolution with a filter comes naturally. Unfortunately, the requirement of a distance (or, at least, of a neighbourhood function) in the input feature space has so far prevented its direct use on data types such as omics data. However, a number of omics data are metrizable, i.e., they can be endowed with a metric structure, enabling to adopt a convolutional based deep learning framework, e.g., for prediction. We propose a generalized solution for CNNs on omics data, implemented through a dedicated Keras layer. In particular, for metagenomics data, a metric can be derived from the patristic distance on the phylogenetic tree. For transcriptomics data, we combine Gene Ontology semantic similarity and gene co-expression to define a distance; the function is defined through a multilayer network where 3 layers are defined by the GO mutual semantic similarity while the fourth one by gene co-expression. As a general tool, feature distance on omics data is enabled by OmicsConv, a novel Keras layer, obtaining OmicsCNN, a dedicated deep learning framework. Here we demonstrate OmicsCNN on gut microbiota sequencing data, for Inflammatory Bowel Disease (IBD) 16S data, first on synthetic data and then a metagenomics collection of gut microbiota of 222 IBD patients., Comment: 7 pages, 3 figures. arXiv admin note: text overlap with arXiv:1709.02268

Generalized Linear Models are routinely used in data analysis. The classical procedures for estimation are based on Maximum Likelihood and it is well known that the presence of outliers can have a large impact on this estimator. Robust procedures are presented in the literature but they need a robust initial estimate in order to be computed. This is especially important for robust procedures with non convex loss function such as redescending M-estimators. Subsampling techniques are often used to determine a robust initial estimate; however when the number of unknown parameters is large the number of subsamples needed in order to have a high probability of having one subsample free of outliers become infeasible. Furthermore the subsampling procedure provides a non deterministic starting point. Based on ideas in Pena and Yohai (1999), we introduce a deterministic robust initial estimate for M-estimators based on transformations Valdora and Yohai (2014) for which we also develop an iteratively reweighted least squares algorithm. The new methods are studied by Monte Carlo experiments.

In this paper, we propose a robust profile estimation method for the parametric and nonparametric components of a single index model when the errors have a strongly unimodal density with unknown nuisance parameter. Under regularity conditions, we derive consistency results for the link function estimators as well as consistency and asymptotic distribution results for the single index parameter estimators. Under a log--Gamma model, the sensitivity to anomalous observations is studied by means of the empirical influence curve. We also discuss a robust $K-$fold procedure to select the smoothing parameters involved. A numerical study is conducted to evaluate the small sample performance of the robust proposal with that of their classical relatives, both for errors following a log--Gamma model and for contaminated schemes. The numerical experiment shows the good robustness properties of the proposed estimators and the advantages of considering a robust approach instead of the classical one.

Background: Convolutional Neural Networks can be effectively used only when data are endowed with an intrinsic concept of neighbourhood in the input space, as is the case of pixels in images. We introduce here Ph-CNN, a novel deep learning architecture for the classification of metagenomics data based on the Convolutional Neural Networks, with the patristic distance defined on the phylogenetic tree being used as the proximity measure. The patristic distance between variables is used together with a sparsified version of MultiDimensional Scaling to embed the phylogenetic tree in a Euclidean space. Results: Ph-CNN is tested with a domain adaptation approach on synthetic data and on a metagenomics collection of gut microbiota of 38 healthy subjects and 222 Inflammatory Bowel Disease patients, divided in 6 subclasses. Classification performance is promising when compared to classical algorithms like Support Vector Machines and Random Forest and a baseline fully connected neural network, e.g. the Multi-Layer Perceptron. Conclusion: Ph-CNN represents a novel deep learning approach for the classification of metagenomics data. Operatively, the algorithm has been implemented as a custom Keras layer taking care of passing to the following convolutional layer not only the data but also the ranked list of neighbourhood of each sample, thus mimicking the case of image data, transparently to the user. Keywords: Metagenomics; Deep learning; Convolutional Neural Networks; Phylogenetic trees, Comment: Presented at BMTL 2017, Naples

A novel approach to obtain weighted likelihood estimates of multivariate location and scatter is discussed. A weighting scheme is proposed that is based on the distribution of the Mahalanobis distances rather than the distribution of the data at the assumed model. This strategy allows to avoid the curse of dimensionality affecting non-parametric density estimation, that is involved in the construction of the weights through the Pearson residuals Markatou et al (1998). Then, weighted likelihood based outlier detection rules and robust dimensionality reduction techniques are developed. The effectiveness of the methodology is illustrated through some numerical studies and real data examples.

Data depth proves successful in the analysis of multivariate data sets, in particular deriving an overall center and assigning ranks to the observed units. Two key features are: the directions of the ordering, from the center towards the outside, and the recognition of a unique center irrespective of the distribution being unimodal or multimodal. This behaviour is a consequence of the monotonicity of the ranks that decrease along any ray from the deepest point. Recently, a wider framework allowing identification of partial centers was suggested in Agostinelli and Romanazzi [2011]. The corresponding generalized depth functions, called local depth functions are able to record local fluctuations and can be used in mode detection, identification of components in mixture models and in cluster analysis. Functional data [Ramsay and Silverman, 2006] are become common nowadays. Recently, Lopez-Pintado and Romo [2011] has proposed the half-region depth suited for functional data and for high dimensional data. Here, following their work we propose a local version of this data depth, we study its theoretical properties and illustrate its behaviour with examples based on real data sets., Comment: 46 pages. 19 figure

robustloggamma is an R package for robust estimation and inference in the generalized loggamma model. We briefly introduce the model, the estimation procedures and the computational algorithms. Then, we illustrate the use of the package with the help of a real data set., Comment: Accepted in Journal of Statistical Software

The generalized log-gamma (GLG) model is a very flexible family of distributions to analyze datasets in many different areas of science and technology. In this paper, we propose estimators which are simultaneously highly robust and highly efficient for the parameters of a GLG distribution in the presence of censoring. We also introduced estimators with the same properties for accelerated failure time models with censored observations and error distribution belonging to the GLG family. We prove that the proposed estimators are asymptotically fully efficient and examine the maximum mean square error using Monte Carlo simulations. The simulations confirm that the proposed estimators are highly robust and highly efficient for finite sample size. Finally, we illustrate the good behavior of the proposed estimators with two real datasets.