An R Package AZIAD for Analyzing Zero-Inflated and Zero-Altered Data
share
Sparse data with a large portion of zeros arise in many scientific disciplines. Modeling sparse data is very challenging due to the skewness of the distribution. We adopt bootstrapped Monte Carlo method to estimate the p-value of the Kolmogorov-Smirnov test, as well as bootstrapped likelihood ratio tests for zero-inflated and zero-altered (or hurdle) model selection. Our new package AZIAD provides miscellaneous functions to simulate zero-inflated or zero-altered data and calculate maximum likelihood estimates of unknown parameters for a large class of discrete or continuous distributions. In addition, we calculate the Fisher information matrix and the confidence intervals of unknown parameters. Compared with other R packages available so far, our package covers many more types of zero-inflated and zero-altered distributions, provides more accurate estimates for unknown parameters, and achieves higher power for model selection. To facilitate the potential users, in this paper we provide theoretical justifications and detailed formulae for functions in AZIAD and illustrate the use of them with executable R code and real dataset.
READ FULL TEXT