Modeling Counts via the Conway-Maxwell-Poisson Distribution: For the Health of It!

• Activity Number: 589
• Type: Topic Contributed
• Date/Time: Thursday, August 7, 2014 : 8:30 AM to 10:20 AM
• Sponsor: Health Policy Statistics Section
• Author(s): Kimberly Sellers*+
• Companies: Georgetown University
• Keywords: over-dispersion ; under-dispersion ; count data

Abstract

Integer-valued data are becoming pervasive in the health sciences (e.g. counting the number of patients with a particular ailment to be admitted to the hospital, or the number of vaccinations to be administered to a particular group). Accordingly, the Poisson distribution is a natural distribution to consider when modeling counts. However, the associated equi-dispersion assumption (i.e. that the mean and variance equal) is constraining. Various alternative models exist to circumvent this issue. In particular, I will introduce the Conway-Maxwell-Poisson (COM-Poisson) distribution as a viable alternative for modeling counts because of its flexibility and ability to capture over- or under-dispersion. Further, I will discuss various statistical methods that have been developed via the COM-Poisson distribution for modeling health data.