Small Numbers Problem

The interpretation of public health data is affected by the problem of interpreting small numbers.

Statisticians and researchers have greater confidence in counts and rates for more common diseases and for larger subpopulations. Epidemiologists tend to have less confidence in rates (count/population) calculated from small populations-at-risk. These rates tend to be unstable, because small changes in the count or in the population lead to wide variance in the rate.

In this tutorial, we will apply two methods to assess and perhaps reduce the influence of small numbers and unstable rates on an analysis. We will:

• mask sparse data by eliminating small-population areas and

• smooth the data to reduce the estimates for small-population areas.