Showing papers by "Francesco Bartolucci published in 2017"

Lmest: An R package for latent Markov models for longitudinal categorical data

Abstract: Latent Markov (LM) models represent an important class of models for the analysis of longitudinal data, especially when response variables are categorical. These models have a great potential of application in many fields, such as economics and medicine. We illustrate the R package LMest that is tailored to deal with the basic LM model and some extended formulations accounting for individual covariates and for the presence of unobserved clusters of units having the same initial and transition probabilities (mixed LM model). The main functions of the package are tailored to parameter estimation through the expectation-maximization algorithm, which is based on suitable forwardbackward recursions. The package also permits local and global decoding and to obtain standard errors for the parameter estimates. We illustrate the use of the package and its main features through some empirical examples in the fields of labour market, health, and criminology.

cquad: An R and Stata Package for Conditional Maximum Likelihood Estimation of Dynamic Binary Panel Data Models

Abstract: We illustrate the R package cquad for conditional maximum likelihood estimation of the quadratic exponential (QE) model proposed by Bartolucci and Nigro (2010) for the analysis of binary panel data. The package also allows us to estimate certain modified versions of the QE model, which are based on alternative parametrizations, and it includes a function for the pseudo-conditional likelihood estimation of the dynamic logit model, as proposed by Bartolucci and Nigro (2012). We also illustrate a reduced version of this package that is available in Stata. The use of the main functions of this package is based on examples using labor market data.

Evaluation of Student Performance through a Multidimensional Finite Mixture IRT Model

Abstract: In the Italian academic system, a student can enroll for an exam immediately after the end of the teaching period or can postpone it; in this second case the exam result is missing. We propose an approach for the evaluation of a student performance throughout the course of study, accounting also for nonattempted exams. The approach is based on an item response theory model that includes two discrete latent variables representing student performance and priority in selecting the exams to take. We explicitly account for nonignorable missing observations as the indicators of attempted exams also contribute to measure the performance (within-item multidimensionality). The model also allows for individual covariates in its structural part.

A Nonparametric Multidimensional Latent Class IRT Model in a Bayesian Framework.

Abstract: We propose a nonparametric item response theory model for dichotomously-scored items in a Bayesian framework. The model is based on a latent class (LC) formulation, and it is multidimensional, with dimensions corresponding to a partition of the items in homogenous groups that are specified on the basis of inequality constraints among the conditional success probabilities given the latent class. Moreover, an innovative system of prior distributions is proposed following the encompassing approach, in which the largest model is the unconstrained LC model. A reversible-jump type algorithm is described for sampling from the joint posterior distribution of the model parameters of the encompassing model. By suitably post-processing its output, we then make inference on the number of dimensions (i.e., number of groups of items measuring the same latent trait) and we cluster items according to the dimensions when unidimensionality is violated. The approach is illustrated by two examples on simulated data and two applications based on educational and quality-of-life data.

Adaptive Quadrature for Maximum Likelihood Estimation of a Class of Dynamic Latent Variable Models

Abstract: Maximum likelihood estimation of models based on continuous latent variables generally requires to solve integrals that are not analytically tractable. Numerical approximations represent a possible solution to this problem. We propose to use the adaptive Gaussian---Hermite (AGH) numerical quadrature approximation for a particular class of continuous latent variable models for time-series and longitudinal data. These dynamic models are based on time-varying latent variables that follow an autoregressive process of order 1, AR(1). Two examples are the stochastic volatility models for the analysis of financial time series and the limited dependent variable models for the analysis of panel data. A comparison between the performance of AGH methods and alternative approximation methods proposed in the literature is carried out by simulation. Empirical examples are also used to illustrate the proposed approach.

Composite likelihood inference in a discrete latent variable model for two-way "clustering-by-segmentation" problems

Abstract: We consider a discrete latent variable model for two-way data arrays, which allows one to simultaneously produce clusters along one of the data dimensions (e.g., exchangeable observational units or features) and contiguous groups, or segments, along the other (e.g., consecutively ordered times or locations). The model relies on a hidden Markov structure but, given its complexity, cannot be estimated by full maximum likelihood. Therefore, we introduce a composite likelihood methodology based on considering different subsets of the data. The proposed approach is illustrated by simulation, and with an application to genomic data.

Granger causality in dynamic binary short panel data models

Abstract: Strict exogeneity of covariates other than the lagged dependent variable, and conditional on unobserved heterogeneity, is often required for consistent estimation of binary panel data models. This assumption is likely to be violated in practice because of feedback effects from the past of the outcome variable on the present value of covariates and no general solution is yet available. In this paper, we provide the conditions for a logit model formulation that takes into account feedback effects without specifying a joint parametric model for the outcome and predetermined explanatory variables. Our formulation is based on the equivalence between Granger's definition of noncausality and a modification of the Sims' strict exogeneity assumption for nonlinear panel data models, introduced by Chamberlain1982 and for which we provide a more general theorem. We further propose estimating the model parameters with a recent fixed-effects approach based on pseudo conditional inference, adapted to the present case, thereby taking care of the correlation between individual permanent unobserved heterogeneity and the model's covariates as well. Our results hold for short panels with a large number of cross-section units, a case of great interest in microeconomic applications.

Job satisfaction and compensating wage differentials: Evidence from Russia

Abstract: This article is focused on the analysis of the determinants of job satisfaction of Russian workers through the estimation of ordered logit models with individual fixed effects on a panel data set extracted from the Russian Longitudinal Monitoring Survey The real wage results positively associated with job satisfaction on all samples considered, after controlling for several time-varying controls aimed at picking up time-varying human capital, preferences, and non-instantaneous equilibrium adjustments As long as the included controls capture all the heterogeneous trends in individual wages, we may interpret this result as a possible failure of the theory of compensating wage differentials in the Russian labour market

Editorial: Special section on latent variable models for longitudinal data.

Abstract: Latent variable (LV) models (see, e.g. Bartholomew et al., 2011) have found an important field of application in the context of life sciences. Their use is justified by the complexity of the biological systems, with implications in terms of sophisticated dependencies between observable variables. In LV models, as it is well known, the observable response variables are affected by (discrete or continuous) variables that are not directly observed. LV models are based on specific assumptions on the conditional distribution of the response variables, given the latent ones. This allows us to model the effect of unobservable covariates (factors) and, thus, to account for the unobserved heterogeneity between subjects. In the last decades, particular attention has been paid to LV models for longitudinal data. In such cases, the phenomenon under investigation evolves over time. This is very common in the field of biometrics and life sciences, where longitudinal studies frequently occur. It is clear that longitudinal designs lead to a more precise and deeper insights into the phenomenon under study. However, they pose problems from a theoretical point of view. Versions of LV models that are specially tailored to the analysis of longitudinal data are based on the assumption that the response variables depend, for instance, on a latent process corresponding to a sequence of continuous latent variables (see, e.g. Skrondal and Rabe-Hesketh, 2004) or on an unobservable Markov chain (see, e.g. Bartolucci et al., 2013). In this way, it is possible to account for time-constant and time-varying unobserved heterogeneity between subjects. Moreover, in longitudinal studies it may occur that only few subjects have complete data records, as individuals participating into the study typically leave it in advance. If this interruption depends on events associated with the outcome of interest, an informative drop-out arises that must be properly modeled. LV models for longitudinal data can also account for this aspect in a sensible way. This Special Section is intended to propose recent advances on LV models for longitudinal data and their applications to real-life studies in the biometrical domain. In the following, we provide a brief introduction to the articles included in this Special Section. The paper by Asar et al. (2016) analyzes data from an an ongoing cohort study on chronic kidney disease. An ad-hoc LV model for longitudinal data is proposed, where the fixed effects are modeled by a piece-wise linear function with three change-points and the random effects are expressed through the sum of a random intercept, a stationary stochastic process with Matérn correlation structure and measurement error. Parameter estimation is carried out through a maximum likelihood approach assuming, for the random effects, a multivariate Normal distribution and a multivariate t distribution, where the latter seems to be more appropriate for the case study considered. To assess the effectiveness of the proposal, the plug-in predictive distributions of the random effects given the data are studied

An Ordinal Latent Markov model for the evaluation of Health Care services

Abstract: This work studies the dynamic behavior of the health status of some elderly hosted in different nursing homes. Specifically, we consider a dataset gathered from the Long Term Care Facilities (LTCF) Programme, a longitudinal study carried on in Umbria (Italy). The final goal of our analysis is to understand whether the evolution of elderly' health conditions significantly change across different nursing homes. To this end, an ordinal Latent Markov model accounting for both dropout and intermittent missing data patterns is proposed. Then, some performance measures are computed on a standardized elderly population in order to rule out the effect of patient case-mix.

A mixture growth model for migrants' remittances: An application to the German Socio-Economic Panel

Abstract: We propose a bivariate mixture growth model with concomitant variables to study the time profiles of international remittances sent by first-generation migrants in Germany from 1996 to 2012. The latent class approach allows us to identify homogeneous sub-groups of migrants associated with different trajectories for their remitting behavior, which can be interpreted in the light of the theoretical economic background. In addition, the latent class model combined with the concomitant variable approach allows us to uncover whether the assignment of migrants to a specific sub-group can be ascribed to their observable characteristics, namely their return intention, as conjectured by the theoretical models.