Effect of Ratio Correlation on Data Interpretation

Abstract

Structural transformation of a model can affect its rationality. Linear transformations of nonlinear models for estimating model parameters often suffer from ratio correlation, the replication of a common variable in the transformed model. A regression analysis of experimental data may indicate a spuriously high correlation, and appear to support the original model, when in fact, it supports the transformed model. Several models commonly used in environmental engineering were analyzed to show the effect of ratio correlation on the estimated model parameters. Since a casual reader often uses the correlation coefficient as a measure of the accuracy of a model, ratio correlation should be avoided in statistical data analysis.