Model is an abstraction of the reality. The first step in data analysis is to capture its most suitable model. This step receives a fatal hit when the data exhibits either over or under variance against the model. This article provides an approach, based on Neyman's C (  ) principle, to decide whether the over or under variance is weak enough for the model to work. To illustrate the method, the times (in hours) between main earthquake and aftershocks during 1973-1995 are considered and explained.