Use of Computer Simulation for Calculating Reference Limits for Distribution Overlap

In laboratory medicine, data collected in different settings or under different conditions are frequently analyzed to obtain meaningful information. Evaluation based only on location or variability measures, although useful, is not enough to extract all the value from data. Other mathematical or statistical approaches are possible, but the specific knowledge required is frequently out of the reach of most laboratorians. Computer simulation may help in solving the problem. The aim of this work was to use computer simulation for calculating reference limits for the overlap of 2 distributions.

Computer simulation was applied to find a reference value, and its confidence limits, from the overlapping area of 2 distributions when population means were allowed to differ by a pre-set amount. The allowable limits were compared with the overlapping area observed between data distributions reported from 3 laboratories. The simulation was run in R language. A description for the experimental setting was added, with the instructions for the simulation reported in structured English.

The simulation allowed estimation of a limit to be used as a reference value when comparing two overlapping areas. Based on these limits, one laboratory in the study was found not to be aligned with the others.

Computer simulation is a low-cost, powerful, and easy to implement tool which may help in solving simple or complex problems, when assumptions are unrealistic or unmet, or when mathematical algorithms are not known or difficult to calculate.