Article

Modelos de regresión para variables expresadas como una proporción continua

Salud publica de Mexico (Impact Factor: 0.94). 01/2006; DOI: 10.1590/S0036-36342006000500006
Source: OAI

ABSTRACT Objetivo. Describir algunas de las alternativas estadísticas disponibles para el estudio de proporciones continuas y comparar los distintos modelos que existen para evidenciar sus ventajas y desventajas, mediante su aplicación a un ejemplo práctico del ámbito de la salud pública. Material y métodos. Con base en la Encuesta Nacional de Salud Reproductiva realizada en el año 2003, se modeló la proporción de cobertura individual en el programa de planificación familiar ¿propuesta en un estudio previo realizado en el Instituto Nacional de Salud Pública en Cuernavaca, Morelos, México (2005)¿ mediante el uso de los modelos de regresión normal, gama, beta y de quasi-verosimilitud. La variante del criterio de información de Akaike (AIC) que propusieron McQuarrie y Tsai se utilizó para definir el mejor modelo. A continuación, y mediante simulación (enfoque Monte Carlo/cadenas de Markov), se generó una variable con distribución beta para evaluar el comportamiento de los cuatro modelos al variar el tamaño de la muestra desde 100 hasta 18 000 observaciones. Resultados. Los resultados muestran que la mejor opción estadística para el análisis de proporciones continuas es el modelo de regresión beta, de acuerdo con sus supuestos y el valor de AIC. La simulación mostró que a medida que aumenta el tamaño de la muestra, el modelo gama y, en especial, el modelo de quasi-verosimilitud se aproximan en grado significativo al modelo beta. Conclusiones. Para la modelación de proporciones continuas se recomienda emplear el enfoque paramétrico de la regresión beta y evitar el uso del modelo normal. Si se tiene un tamaño de muestra grande, el uso del enfoque de quasiverosimilitud representa una buena alternativa.

0 Bookmarks
 · 
127 Views
  • Source
    [Show abstract] [Hide abstract]
    ABSTRACT: We discuss some practical aspects of fitting and interpreting nonlinear quasi-likelihood regression models. The examples considered are for data in the form of continuous rates or proportions distributed on the unit interval. However, many of the concepts and procedures illustrated should be helpful with other models as well. Our approach to estimation for multi-parameter models is by the method of fitting expected values discussed by Jennrich and Moore (1975) and Cox (1984a). This allows the use of standard nonlinear regression computer programs. An important feature of this approach is that it does not require the existence of a link function. Thus nonlinear models are easily accommodated, including models with a parametric link function, as well as models with a link to a nonlinear predictor having multiplicative terms. After a brief discussion of the properties of these models and the method of fitting, we provide a detailed illustration of fitting a number of link-linear as well as nonlinear models to a set of data considered by Wedderburn (1974). The analysis makes use of a diagnostic tool known as the biplot to help select a multiplicative interaction model. A second example of a useful nonlinear model is provided by an experiment in toxicology. As part of the discussion of these alternative models, we illustrate the choice of link and variance functions using an extended quasi-likelihood developed by Nelder and Pregibon (1987).
    Computational Statistics & Data Analysis. 01/1996;
  • [Show abstract] [Hide abstract]
    ABSTRACT: We show that if firms locate on a circle according to a uniform distribution, their market shares will be beta distributed. The result implies that concentration ratios will be skewed. We apply the result to the market for auditing services in the US and show the beta distribution outperforms other relevant distributions.
    Review of Industrial Organization 11/1995; 10(6):737-747. · 0.48 Impact Factor
  • Source
    [Show abstract] [Hide abstract]
    ABSTRACT: Many types of studies examine the inè uence of selected variables on the conditional expectation of a proportion or vector of proportions, for example, market shares, rock composition, and so on. We identify four distributional categories into which such data can be put, and focus on regression models for the é rst category, for proportions observed on the open interval (0, 1). For these data, we identifydifferent specié cations used in prior research and compare these specié cations using two common samples and specié cations of the regressors. Based upon our analysis, we recommend that researchers use either a parametric regression model based upon the beta distribution or a quasi-likelihood regression model developed by Papke and Wooldridge (1997) for these data. Concerning the choice between these two regression models, we recommend that researchers use the parametric regression model unless their sample size is large enough to justify the asymptotic arguments underlying the quasi-likelihood approach.
    Statistical Modelling 01/2003; 3(3):193-213. · 0.68 Impact Factor

Full-text

View
1 Download
Available from