Seminar on Proximity Data and Multivariate Analytics

Fecha: Lunes 8 de febrero de 2016
Lugar: Sala de Conferencias de la Facultad de Ciencias
Horario: de 10 a 14
Programa:
- Albert Satorra (Universitat Pompeu Fabra. Barcelona): Univariate versus multivariate modelling of panel data, specification issues and goodness of fit testing
  - Abstract: Panel data can be arranged into a matrix in two ways, called `long’ and `wide’ formats (LF and WF). The two formats suggest two alternative model approaches for analyzing panel data: (i) univariate regression with varying intercept; and (ii) multivariate regression with latent variables (a particular case of structural equation model, SEM). The present paper compares the two approaches showing in which circumstances they yield equivalent in some cases, even numerically equal results. We show that the univariate approach gives results equivalent to the multivariate approach when restrictions of time invariance (in the paper, the TI assumption) are imposed on the parameters of the multivariate model. The multivariate perspective we offer makes visible the implicit restrictions implied by the univariate approach and reinterprets a classical in the econometric literature of fixed versus random effects model as identification issues in a structural equation model (SEM) framework. The setting side by side the univariate versus multivariate perspectives to panel data shows the easy way of the multivariate approach to tackle the dynamics into the model. It is shown that the restrictions implicit in the univariate approach can be assessed by chi-square difference testing of two nested multivariate models. In addition, common tests encountered in the econometric analysis of panel data, such as the Hausman test, are shown to have an equivalent representation as chi-square difference tests. Commonalities and differences between the univariate and multivariate approaches are illustrated using an empirical panel data set of firms’ profitability as well as a simulated panel data.
- Daniel Oberski (Tiburg University, The Netherlands): A measure to evaluate model fit by sensitivity analysis.
  - Abstract: Latent variable models involve restrictions on the data that can be formulated in terms of «misspecifications»: restrictions with a model-based meaning. Examples include zero cross-loadings and local dependencies, as well as “measurement invariance” or “differential item functioning”. If incorrect, misspecifications can potentially disturb the main purpose of the latent variable analysis—seriously so in some cases. I propose to evaluate whether a particular analysis at hand is such a case or not. To do this, I define a measure based on the likelihood of the restricted model that approximates the change in the parameters of interest if the misspecification were freed, the EPC-interest. The main idea is to examine the EPC-interest and free those misspecifications that are “important” while ignoring those that are not. I have implemented the EPC-interest in the lavaan software for structural equation modeling and the Latent Gold software for latent class analysis. This approach can resolve several problems and inconsistencies in the current practice of model fit evaluation used in latent variable analysis, something I illustrate using analyses from the “measurement invariance” literature and from item response theory.
- Antonio D’Ambrosio (Università degli Studi di Napoli Federico II, Italy): A recursive partitioning method for the prediction of preference rankings based upon Kemeny distances
  - Abstract: Preference rankings usually depend on the characteristics of both the individuals judging a set of objects and the objects being judged. This topic has been handled in the literature with log-linear representations of the generalized Bradley-Terry model and, recently, with distance-based tree models for rankings. A limitation of these approaches is that they only work with full rankings or with a pre-specified pattern governing the presence of ties, and/or they are based on quite strict distributional assumptions. To overcome these limitations, we propose a new prediction tree method for ranking data that is totally distribution-free. It combines Kemeny’s axiomatic approach to define a unique distance between rankings with the CART approach to find a stable prediction tree. Furthermore, our method is not limited by any particular design of the pattern of ties. The method is evaluated in an extensive full-factorial Monte Carlo study with a new simulation design
Organiza: Programa de Doctorado en Estadística Matemática y Aplicada
Más información: José Fernando Vera
Departamento de Estadística e I.O.
Facultad de Ciencias
Universidad de Granada
18071 Granada
SPAIN
Tlf: +34958243157
Fax: +34958243267
jfvera@ugr.es

Seminar on Proximity Data and Multivariate Analytics

08 febrero 2016 | 09:00 - 13:00 h

Actividades culturales y de divulgación
Conferencias, seminarios, divulgación científica

Fechas y horarios

Seminar on Proximity Data and Multivariate Analytics

08 febrero 2016 | 09:00 - 13:00 h

Actividades culturales y de divulgaciónConferencias, seminarios, divulgación científica

Fechas y horarios

Actividades culturales y de divulgación
Conferencias, seminarios, divulgación científica