부산대학교 도서관

Count data with a single mode at zero and a few extreme counts in the outcome variable are frequently seen in medical research. In this article, the response variable—the number of doctor visits in a 3-month interval—exhibited an inflated zero frequency count and a few extreme values. We introduce a right-censored zero-inflated Conway–Maxwell Poisson (CZICMP) regression model to simultaneously handle the problems of excess zeros and extreme values. Parameter estimates were obtained using maximum likelihood. We compared the CZICMP model to alternative count regression models and assessed performance via a case study and a simulation study. [ABSTRACT FROM AUTHOR]