Is Poisson Dispersion Diluted or Over-Saturated? An Index is Created to Answer

Abstract

A prelude to interpret a pattern in the repeating incidences is to identify the underlying frequency distribution of the collected data. A case in point is the Poisson distribution which is often selected for medical count data such as gene mutations, medication error and number of ambulatory pickups in a day. A requirement for the Poisson distribution is that the variance ought to be equal to the mean. The variance signifies the volatility in the occurrences. An implication is that the volatility becomes more when the average incidence is higher. When this requirement of the functional equivalence of the Poisson mean and variance is breached, the data deviates from a Poisson distribution. How could a data analyst recognize and point out to the medical team the dilution level of the requirement in their data? For this purpose, a simple and easier geometrical approach is developed in this article and illustrated with several historical data sets in the literature.