Paquete ZOIP de R para analizar datos proporcionales inflados con ceros y/o unos

• Autores: Juan Camilo Diaz Zapata, Freddy Hernández Barajas, Olga C. Úsuga-Manco
• Localización: Comunicaciones en Estadística, ISSN 2027-3355, ISSN-e 2339-3076, Vol. 11, Nº. 1, 2018, págs. 37-61
• Idioma: español
• DOI: 10.15332/s2027-3355.2018.0001.02
• Títulos paralelos:
  • ZOIP package in R to analyse inflated proportional data with zeros and/or ones
• Enlaces
  • Texto completo (pdf)
• Resumen
  • español

    Los datos obtenidos a partir de variables medidas bajo porcentajes, tasas y proporciones, son llamados datos proporcionales y estos se encuentran ubicados por lo general en el intervalo (0,1). Diferentes distribuciones se han desarrollado para caracterizar este tipo de variables, sin embargo, existe la posibilidad que este tipo de variables puedan dar resultados de cero o uno. Haciendo que se implementaran dos distribuciones beta inflada con ceros y/o unos, diferenciadas solo por su parametrización. Otros autores han trabajo con otras distribuciones para datos proporcionales inflados, como la distribución simplex, sin embargo, no se ha encontrado una distribución que reúna las principales distribuciones para datos proporcionales inflados con ceros y/o unos, en este trabajo se presenta el paquete ZOIP del sistema de computación R para la implementación de la distribución ZOIP (Zeros Ones Inflated Proportional) que reúne la distribución simplex y beta bajo diferentes parametrizaciones y estima los parámetros de dicha distribución. La estimación de los parámetros de la distribución ZOIP se hace vía máxima verosimilitud. Se realizan estudios de simulación que muestran la convergencia satisfactoria de los parámetros y se presenta el ajuste de una distribución ZOIP a datos reales.

  • English

    Data obtained from variables measured as percentages, rates and proportions are called proportional data and these variables are usually within the (0,1) interval. Although, different distributions have been developed to characterize these variables, it is possible that some variables take the extreme values at zero and one. Some researchers as Ospina & Ferrari (2012) and Rigby & Stasinopoulos (2005) proposed the inflated Beta distribution with zeros and/or ones, differentiated only by their parameterization. Other authors such as Galvis & Lachos (2014) have worked with inflated proportional data, considering other distributions such as simplex distributions. Nevertheless, it has not been found any distribution which combines the main characteristic of several distributions for solving these issue. Thus, this paper presents the ZOIP package (Zeros Ones Inflated Proportional), developed on the computation system R, this package combines both Beta and simplex distribution and estimates its parameters given the parameters for the original distributions. The parameters estimation process, is performed by maximum likelihood method. Finally, simulations studies have been performed which show the accuracy of the parameters convergence and the adjustment of a ZOIP distribution on real data

• Referencias bibliográficas
  • Ferrari, S. & Cribari-Neto, F. (2004), ‘Beta regression for modelling rates and proportions’, Journal of applied statistics 31(7), 799–815.
  • Galvis, D. & Lachos, V. (2014), Bayesian analysis of regression models for proportional data in the presence of zeros and ones, PhD thesis,...
  • Hahn, E. (2008), ‘Mixture densities for project management activity times: A robust approach to pert’, European Journal of operational research...
  • Hossain, A. (2015), ‘Modelling a proportion response variable using generalized additive models for location scale and shape’, International...
  • Jørgensen, B. (1997), ‘The theory of dispersion models’, Computational statistics and Data analysis 76.
  • Jørgensen, B. & Barndorff-Nielsen (1991), ‘Some parametric models on the simplex’, Journal of multivariate analysis 39(1), 106–116.
  • Kosmidis, I., Zou, A. & Jagan, K. (2015), ‘Beyond beta regression: modelling percentages and fractions in the presence of boundary observations’,...
  • Ospina, R. & Ferrari, S. (2010), ‘Inflated beta distributions online: doi:10.1007/s00362-008-0125-4’, Statitical papers .
  • Ospina, R. & Ferrari, S. (2012), ‘A general class of zero-or-one inflated beta regression models’, Computational Statistics and Data Analysis...
  • Owen, C. (2008), Parameter estimation for the beta distribution, Master’s thesis, Brigham young university.
  • Pace, L. & Salvan, A. (1997), ‘Principles of statistical inference from a neo-fisherian perspective. in: Advanced series on statistical...
  • Qiu, Z., Song, P. & Tan, M. (2008), ‘Simplex mixed-effects models for longitudinal proportional data’, Scandinavian Journal of Statistics...
  • R Core Team (2017), R: A Language and Environment for Statistical Computing, R Foundation for Statistical Computing, Vienna, Austria. *http://www.R-project.org/
  • Rigby, B. & Stasinopoulos, M. (2005), ‘Generalized additive models for location, scale and shape’, Applied Statistical 54(3), 507–554.
  • Rigby, B. & Stasinopoulos, M. (2008), ‘Instructions on how to use the gamlss package in r’, Computational statistics and Data analysis...
  • Zeileis, A., Cribari-Neto, F. & Gr¨un, B. (2010), ‘Beta regression in r’, Journal of statistical software 34(2), 1–24.
  • Zhang, P., Qiu, Z. & Shi, C. (2016), ‘simplexreg: An r package for regression analysis of proportional data using the simplex distribution’,...