Extended zero-one inflated beta and adjusted three-part regression models for proportional data analysis

ABSTRACT

Extended zero-one inflated beta and adjusted three-part regression models are introduced to analyze proportional response data where there are nonzero probabilities that the response variable takes the values zero and one. The proposed models adapt skewness and heteroscedasticity of the fractional response data, and are constructed to estimate the unknown parameters. Extensive Monte Carlo simulation studies are used to compare the performance of the two approaches with respect to bias and root mean square error. A real data example is presented to illustrate the application of both regression models.

KEYWORDS:

• Adjusted three-part model
• extended zero-one inflated beta model
• Monte Carlo study
• proportional response data
• root mean square error

MATHEMATICS SUBJECT CLASSIFICATION:

• 62F10