Laura Trinchera

Significant correlation between a set of genetic polymorphisms and a functional brain network revealed by feature selection and sparse Partial Least Squares

Brain imaging is increasingly recognised as an intermediate phenotype to understand the complex path between genetics and behavioural or clinical phenotypes. In this context, a first goal is to propose methods to identify the part of genetic variability that explains some neuroimaging variability. Classical univariate approaches often ignore the potential joint effects that may exist between genes or the potential covariations between brain regions. In this paper, we propose instead to investigate an exploratory multivariate method in order to identify a set of Single Nucleotide Polymorphisms (SNPs) covarying with a set of neuroimaging phenotypes derived from functional Magnetic Resonance Imaging (fMRI). Recently, Partial Least Squares (PLS) regression or Canonical Correlation Analysis (CCA) have been proposed to analyse DNA and transcriptomics. Here, we propose to transpose this idea to the DNA vs. imaging context. However, in very high-dimensional settings like in imaging genetics studies, such multivariate methods may encounter overfitting issues. Thus we investigate the use of different strategies of regularisation and dimension reduction techniques combined with PLS or CCA to face the very high dimensionality of imaging genetics studies. We propose a comparison study of the different strategies on a simulated dataset first and then on a real dataset composed of 94 subjects, around 600,000 SNPs and 34 functional MRI lateralisation indexes computed from reading and speech comprehension contrast maps. We estimate the generalisability of the multivariate association with a cross-validation scheme and demonstrate the significance of this link, using a permutation procedure. Univariate selection appears to be necessary to reduce the dimensionality. However, the significant association uncovered by this two-step approach combining univariate filtering and L1-regularised PLS suggests that discovering meaningful genetic associations calls for a multivariate approach.

New Perspectives in Partial Least Squares and Related Methods

New Perspectives in Partial Least Squares and Related Methods shares original, peer-reviewed research from presentations during the 2012 partial least squares methods meeting (PLS 2012). This was the 7th meeting in the series of PLS conferences and the first to take place in the USA. PLS is an abbreviation for Partial Least Squares and is also sometimes expanded as projection to latent structures. This is an approach for modeling relations between data matrices of different types of variables measured on the same set of objects. The twenty-two papers in this volume, which include three invited contributions from our keynote speakers, provide a comprehensive overview of the current state of the most advanced research related to PLS and related methods. Prominent scientists from around the world took part in PLS 2012 and their contributions covered the multiple dimensions of the partial least squares-based methods. These exciting theoretical developments ranged from partial least squares regression and correlation, component based path modeling to regularized regression and subspace visualization. In following the tradition of the six previous PLS meetings, these contributions also included a large variety of PLS approaches such as PLS metamodels, variable selection, sparse PLS regression, distance based PLS, significance vs. reliability, and non-linear PLS. Finally, these contributions applied PLS methods to data originating from the traditional econometric/economic data to genomics data, brain images, information systems, epidemiology, and chemical spectroscopy. Such a broad and comprehensive volume will also encourage new uses of PLS models in work by researchers and students in many fields.

Role of intrinsic and extrinsic factors in user social media acceptance within workspace: Assessing unobserved heterogeneity

Abstract This study develops and empirically tests a theoretical extension of a technology acceptance model that integrates intrinsic and extrinsic motivators into IT acceptance to predict the adoption of social media within the workspace. The model was tested using cross-sectional data collected from different workplaces in different geographic regions. To detect the homogeneity of users’ behavior, we used a response-based procedure for partial least squares. The model was strongly supported for the global model. Our results revealed the existence of distinct adoption behaviors for different groups within the overall sample. These findings advance theory and contribute to future research on social media adoption.

Globally Sparse PLS Regression

Partial least squares (PLS) regression combines dimensionality reduction and prediction using a latent variable model. It provides better predictive ability than principal component analysis by taking into account both the independent and response variables in the dimension reduction procedure. However, PLS suffers from over-fitting problems for few samples but many variables. We formulate a new criterion for sparse PLS by adding a structured sparsity constraint to the global SIMPLS optimization. The constraint is a sparsity-inducing norm, which is useful for selecting the important variables shared among all the components. The optimization is solved by an augmented Lagrangian method to obtain the PLS components and to perform variable selection simultaneously. We propose a novel greedy algorithm to overcome the computation difficulties. Experiments demonstrate that our approach to PLS regression attains better performance with fewer selected predictors.

Dimension Reduction and Regularization Combined with Partial Least Squares in High Dimensional Imaging Genetics Studies

In the imaging genetics field, the classical univariate approach ignores the potential joint effects between genes or the potential covariations between brain regions. In this paper, we propose instead to investigate exploratory multivariate methods, namely partial least squares regression or canonical correlation analysis, in order to identify a set of genetic polymorphisms covarying with a set of neuroimaging phenotypes. However, in high-dimensional settings, such multivariate methods may encounter overfitting issues. Thus, we investigate the use of different strategies of regularization and dimension reduction, combined with PLS or CCA, to face the very high dimensionality of imaging genetics studies. We propose a comparison study of the different strategies on a simulated dataset. We estimate the generalisability of the multivariate association with a cross-validation scheme and assess the capacity of good detection. Univariate selection seems necessary to reduce the dimensionality. However, the best results are obtained by combining univariate filtering and L 1-regularized PLS, which suggests that discovering meaningful genetic associations calls for a multivariate approach.

A Distribution Free Interval Estimate for Coefficient Alpha

Scales are important tools for obtaining quantitative measures of theoretical constructs. Once a set of measures to be used in a scale is selected, reliability is commonly examined in order to assess their measurement quality. To date, Cronbach’s coefficient alpha is the most commonly reported index of measurement quality for assessing scale reliability. In this paper, an asymptotic distribution of the natural estimator of coefficient alpha is derived. A new interval estimate and a statistical test on the significance of the sample estimate of the coefficient are also presented. The proposed approach is compared to four popular methods commonly used to compute confidence intervals (CI) for alpha using a Monte Carlo simulation study. An R function for implementing the proposed CI approach is also provided.

Evaluation de la présence des capacités marketing: Proposition d’un index multidimensionnel et hiérarchique

Notre ambition est de proposer un instrument multidimensionnel permettant de décrire le degré de présence des principales capacités marketing sur trois niveaux d’abstraction. Après avoir présenté le cadre théorique relatif aux capacités marketing, l’article souligne tout d’abord les limites des principales échelles proposées par Vorhies et al. (1999 ; 2009), Vorhies et Harker (2000), et Vorhies et Morgan (2003 ; 2005). Ensuite, les étapes nécessaires au développement et à la validation d’un index multidimensionnel formatif de troisième ordre sont détaillées. Sur la base d’une collecte de données réalisée auprès d’un échantillon de 199 PME françaises, la phase d’analyse de la validité convergente et discriminante de l’instrument est réalisée à l’aide de l’approche PLS aux modèles à variables latentes (PLS-PM). Enfin, la validité nomologique de l’instrument proposé est confirmée via l’étude de l’influence des capacités marketing sur la performance organisationnelle.

Evaluating the presence of marketing capabilities: A multidimensional, hierarchical index

We propose a multidimensional instrument to assess the degree of presence of marketing capabilities a firm possesses, at three levels of abstraction. We first present the theoretical framework for marketing capabilities and discuss the main scales proposed by Vorhies et al. Then, we detail the steps required to develop and validate our third-order formative instrument. We assess the convergent and discriminant validity of the proposed instrument via partial least squares path modelling (PLS-PM) applied to a sample of 199 French small- and medium-sized enterprises (SMEs). Finally, we check the nomological validity of our instrument by testing the positive effect of marketing capabilities on organisational performance.

Integrating Non-metric Data in Partial Least Squares Path Models: Methods and Application

In this chapter we discuss how to include non-metric variables (i.e., ordinal and/or nominal) in a PLS Path Model. We present the Non-Metric PLS approach for handling these type of variables, and we integrate the logistic regression into the PLS Path model for predicting binary outcomes. We discuss features and properties of these PLS Path Modeling enhancements via an application on real data. We use data collected by merging the archives of Sapienza University of Rome and the Italian Ministry of Labor and Social Policy. The analysis of this data measured quantitatively, for the first time in Italy, the impact of graduates’ Educational Performance on the first 3 years of their job career.

Discovering associations in high dimensional imaging-genetics data: A comparison study of dimension reduction and regularisation strategies combined with partial least squares

Brain imaging is increasingly recognised as an intermediate pheno-type in the understanding of the complex path between genetics and behavioural or clinical phenotypes. In this context, a first goal is to propose methods to identify the part of genetic variability that explains some neuroimaging variability. Here, we investigate multi-variate methods, Partial Least Squares (PLS) regression and Canonical Correlation Analysis (CCA), in order to identify a set of Single Nucleotide Polymorphisms (SNPs) covarying with a set of neuroimaging phenotypes derived from functional Magnetic Resonance Imaging (fMRI). Because in such high-dimensional settings multi-variate methods overfit the data, we propose a comparison study of several dimension reduction and regularisation strategies combined with PLS or CCA. We demonstrate that the combination of univariate filtering and sparse PLS outperforms all other strategies and is able to extract a significant link between a set of SNPs and a set of brain regions activated during a reading task.