This thesis which consists of four papers is concerned with estimation methods in factor analysis and structural equation models. New estimation methods are proposed and investigated. In paper I an approximation of the penalized maximum likelihood (ML) is introduced to fit an exploratory factor analysis model. Approximated penalized ML continuously and efficiently shrinks the factor loadings towards zero. It naturally factorizes a covariance matrix or a correlation matrix. It is also applicable to an orthogonal or an oblique structure. Paper II, a simulation study, investigates the properties of approximated penalized ML with an orthogonal factor model. Different combinations of penalty terms and tuning parameter selection methods are examined. Differences in factorizing a covariance matrix and factorizing a correlation matrix are also explored. It is shown that the approximated penalized ML frequently improves the traditional estimation-rotation procedure. In Paper III we focus on pseudo ML for multi-group data. Data from different groups are pooled and normal theory is used to fit the model. It is shown that pseudo ML produces consistent estimators of factor loadings and that it is numerically easier than multi-group ML. In addition, normal theory is not applicable to estimate standard errors. A sandwich-type estimator of standard errors is derived. Paper IV examines properties of the recently proposed polychoric instrumental variable (PIV) estimators for ordinal data through a simulation study. PIV is compared with conventional estimation methods (unweighted least squares and diagonally weighted least squares). PIV produces accurate estimates of factor loadings and factor covariances in the correctly specified confirmatory factor analysis model and accurate estimates of loadings and coefficient matrices in the correctly specified structure equation model. If the model is misspecified, robustness of PIV depends on model complexity, underlying distribution, and instrumental variables.

Considering 91 countries with seaports, this study conducted an empirical inquiry into the broader economic contribution of seaborne trade, from a port infrastructure quality and logistics performance perspective. Investment in quality improvement of port infrastructure and its contribution to economy are often questioned by politicians, investors and general public. A structural equation model (SEM) is used to provide empirical evidence of significant economic impacts of port infrastructure quality and logistics performance. Furthermore, analysis of a multi-group SEM is performed by dividing countries into developed and developing economy groups. The results reveal that it is vital for developing countries to continuously improve the quality of port infrastructure as it contributes to better logistics performance, leading to higher seaborne trade, yielding higher economic growth. However, this association weakens as the developing countries become richer.

Ce travail concerne les analyses visant à étudier les données où les individus sont structurés en différents groupes (données multi-groupes). La thèse aborde la question des données multi-groupes ayant une structure en un seul tableau, plusieurs tableaux, trois voies et deux blocs (régression). Cette thèse présente plusieurs méthodes d'analyse de données multi-groupes dans le cadre de l'analyse factorielle. Notre travail comporte trois parties. La première partie traite de l'analyse de données multi-groupes (un bloc de variables divisé en sous-groupes d'individus). Le but est soit descriptif (analyse intra-groupes) ou prédictif (analyse discriminante ou analyse inter-groupe). Nous commençons par une description exhaustive des méthodes multi-groupes. En outre, nous proposons deux méthodes : l'Analyse Procrustéenne duale et l'Analyse en Composantes Communes et Poids Spécifiques duale. Nous exposons également de nouvelles propriétés et algorithmes pour l'Analyse en Composantes Principales multi-groupes. La deuxième partie concerne l'analyse multi-blocs et multi-groupes et l'analyse trois voies et multi-groupes. Nous présentons les méthodes existantes. Par ailleurs, nous proposons deux méthodes, l'ACP multi-blocs et multi-groupes et l'ACP multi-blocs et multi-groupes pondérée, vues comme des extensions d'Analyse en Composantes Principales multi-groupes. L'analyse en deux blocs et multi-groupes est prise en compte dans la troisième partie. Tout d'abord, nous présentons des méthodes appropriées pour trouver la relation entre un ensemble de données explicatives et un ensemble de données à expliquer, les deux tableaux présentant une structure de groupe entre les individus. Par la suite, nous proposons quatre méthodes pouvant être vues comme des extensions de la régression PLS au cas multi-groupes, et parmi eux, nous en sélectionnons une et la développons dans une stratégie de régression. Les méthodes proposées sont illustrées sur la base de plusieurs jeux de données réels dans le domaine de la biologie. Toutes les stratégies d'analyse sont programmées sur le logiciel libre R. / This work deals with multi-group analysis, to study multi-group data where individuals are a priori structured into different groups. The thesis tackles the issue of multi-group data in a multivariate, multi-block, three-way and two-block (regression) setting. It presents several methods of multi-group data analysis in the framework of factorial analysis. It includes three sections. The first section concerns the case of multivariate multi-group data. The aim is either descriptive (within-group analysis) or predictive (discriminant analysis, between-group analysis). We start with a comprehensive review of multi-group methods. Furthermore, we propose two methods namely Dual Generalized Procrustes Analysis and Dual Common Component and Specific Weights Analysis. We also exhibit new properties and algorithms for multi-group Principal Component Analysis. The second section deals with multiblock multi-group and three-way multi-group data analysis. We give a general review of multiblock multi-group methods. In addition, we propose two methods, namely multiblock and multi-group PCA and Weighted-multiblock and multi-group PCA, as extensions of multi-group Principal Component Analysis. The two-block multi-group analysis is taken into account in the third section. Firstly, we give a presentation of appropriate methods to investigate the relationship between an explanatory dataset and a dependent dataset where there is a group structure among individuals. Thereafter, we propose four methods, namely multi-group PLS, in the PLS approach, and among them we select one and develop it into a regression strategy. The proposed methods are illustrated on the basis of several real datasets in the field of biology. All the strategies of analysis are implemented within the framework of R.