Análisis de un modelo de ecuaciones estructurales para el consumo de drogas en Costa Rica 2015
Loading...
Date
Authors
Obando Villalobos, Leidy Priscila
Journal Title
Journal ISSN
Volume Title
Publisher
Abstract
El fenómeno de las drogas es un tema complejo y asimismo los factores que llevan a una persona a comenzar a consumir drogas y las razones por las cuales se desarrolla el trastorno de dependencia, se relacionan con una serie de determinantes estructurados en diferentes niveles del entorno del individuo (biológicos, psicológicos y sociales). En este estudio se emplean los modelos de ecuaciones estructurales para analizar los patrones de consumo de drogas tomando en cuenta características de los sujetos y de su entorno. El estudio se enfocó en evaluar las relaciones causales entre los factores latentes y analizar el efecto moderador de la edad y el sexo sobre estas. Se compararon los resultados obtenidos con el estimador WLS y el enfoque bayesiano con un modelo no informativo y otro modelo con distribuciones previas para las cargas factoriales como Normales con media cero y variancia 1. De los modelos analizados únicamente el modelo WLS presentó un buen ajuste aunque no se encontró evidencia significativa que apoyara las hipótesis planteadas en este estudio en cuanto a las relaciones causales entre los determinantes individuales, el patrón de consumo y el consumo problemático, sin embargo, se señala que los hallazgos sugieren que pueden existir problemas de medición así como otras variables no incluidas en el análisis que podrían afectar la estimación y la significancia de los coeficientes. Al incorporar el enfoque multigrupo, los modelos frecuentistas no fueron estimables y los problemas de ajuste persistieron en los modelos bayesianos, por lo que no se pudieron valorar las hipótesis de invariancia que refieren al efecto moderador del sexo y la edad.
The phenomenon of drug use is a complex issue, and likewise, the factors that lead a person to start using drugs and the reasons why they develop dependence disorder are related to a series of determinants structured at different levels of the individual's environment (biological, psychological, and social). In this study, structural equation models are used to analyze drug use patterns, taking into account both individual and environmental characteristics. The study focused on evaluating the causal relationships between latent factors and analyzing the moderating effect of age and sex on these relationships. The results obtained with the WLS estimator and the Bayesian approach with a non-informative model and another model with prior distributions for the factorial loadings as Normal with mean zero and variance 1 were compared. Of the models analyzed, only the WLS model showed a good fit, although no significant evidence was found to support the hypotheses proposed in this study regarding the causal relationships between individual determinants, consumption patterns, and problematic consumption. However, it is noted that the findings suggest that there may be measurement issues as well as other variables not included in the analysis that could affect the estimation and significance of the coefficients. By incorporating the multigroup approach, the frequentist models were not estimable, and the fit problems persisted in the Bayesian models, so the hypotheses of invariance referring to the moderating effect of sex and age could not be assessed.
The phenomenon of drug use is a complex issue, and likewise, the factors that lead a person to start using drugs and the reasons why they develop dependence disorder are related to a series of determinants structured at different levels of the individual's environment (biological, psychological, and social). In this study, structural equation models are used to analyze drug use patterns, taking into account both individual and environmental characteristics. The study focused on evaluating the causal relationships between latent factors and analyzing the moderating effect of age and sex on these relationships. The results obtained with the WLS estimator and the Bayesian approach with a non-informative model and another model with prior distributions for the factorial loadings as Normal with mean zero and variance 1 were compared. Of the models analyzed, only the WLS model showed a good fit, although no significant evidence was found to support the hypotheses proposed in this study regarding the causal relationships between individual determinants, consumption patterns, and problematic consumption. However, it is noted that the findings suggest that there may be measurement issues as well as other variables not included in the analysis that could affect the estimation and significance of the coefficients. By incorporating the multigroup approach, the frequentist models were not estimable, and the fit problems persisted in the Bayesian models, so the hypotheses of invariance referring to the moderating effect of sex and age could not be assessed.
Description
Keywords
farmacodependencia, estadística