Nazmin Akter * and Rezaul Karim
Count data are now extensively available in a wide range of disciplines. The Poisson distribution, the most used for modeling count data, assumes equidispersion (variance and mean are equal). Poisson models are less suitable for modeling since observed count data frequently display under dispersion or over dispersion. To handle a variety of dispersion levels alternative regression models including negative binomial regression, generalized Poisson regression, and most recently Conway Maxwell-Poisson (COM-Poisson) regression models are employed. Using dispersed data; we compared the COM-Poisson to all other regression models and illustrated how effective and better it is. We conducted a case study utilizing COVID-19 daily death data related to meteorological factors to show how models are applied to real domains.
PDFShare this article
Journal of Biometrics & Biostatistics received 3496 citations as per Google Scholar report