A new regression model for rates and proportions data with applications.

Autor: Prataviera F; Department Exact Sciences, University of São Paulo, Piracicaba, SP, Brazil., Cordeiro GM; Department Exact Sciences, University of São Paulo, Piracicaba, SP, Brazil., Ortega EMM; Department Exact Sciences, University of São Paulo, Piracicaba, SP, Brazil., Hashimoto EM; Academic Department of Mathematics - Federal University Technological of Paraná, Paraná, PR, Brazil., Cancho VG; Department of Applied Mathematics and Statistics, University of São Paulo, São Carlos, SP, Brazil.
Jazyk: angličtina
Zdroj: Journal of applied statistics [J Appl Stat] 2021 Sep 03; Vol. 49 (16), pp. 4137-4161. Date of Electronic Publication: 2021 Sep 03 (Print Publication: 2022).
DOI: 10.1080/02664763.2021.1973385
Abstrakt: We propose a new continuous distribution in the interval ( 0 , 1 ) based on the generalized odd log-logistic-G family, whose density function can be symmetrical, asymmetric, unimodal and bimodal. The new model is implemented using the gamlss packages in R. We propose an extended regression based on this distribution which includes as sub-models some important regressions. We employ a frequentist and Bayesian analysis to estimate the parameters and adopt the non-parametric and parametric bootstrap methods to obtain better efficiency of the estimators. Some simulations are conducted to verify the empirical distribution of the maximum likelihood estimators. We compare the empirical distribution of the quantile residuals with the standard normal distribution. The extended regression can give more realistic fits than other regressions in the analysis of proportional data.
Competing Interests: No potential conflict of interest was reported by the author(s).
(© 2021 Informa UK Limited, trading as Taylor & Francis Group.)
Databáze: MEDLINE
Externí odkaz:
Nepřihlášeným uživatelům se plný text nezobrazuje