Variation is a second-order stereological parameter.

The primary goal of unbiased stereology is to make sample estimates of first-order population parameters, e.g., number, volume, length, and surface area; and, to estimate their variation.

Two important benefits arise when bias is successfully eliminated from stereological designs.

1. Accuracy - analysis of more individuals causes sample estimates to converge on the expected value of the population parameter.
2. Partitioning observed variance into its random error, which consists of biological variance and sampling variance (error). The unbiasedness of the methods allows for optimization of sampling for maximum efficiency, and thus minimizing time, labor, and material expenses.

Typical observed variance in a sample estimate will be high when less than five individuals are analyzed. In this case, statistical power is low, i.e., there are too few individuals to make a reliable estimate. Before adding more individuals to the analysis, which can be costly in terms of time, labor, and material, the first step is to optimize the sampling protocol for maximum efficiency.

Optimization allows you to ensure that the majority of variation in the sample estimate is from biological variation, rather than sampling variation.

Because we are making estimates based on sampling only a part of the total tissue of interest, we are in control of the sampling intensity. By partitioning the observed variance into its component sources, we can identify the source of variance that is making the greatest contribution to the observed variance.

The two components of total variance in a sample estimate are biological variance and sampling variance (error).