A comparison of ratio-based and covariance analyses of a nutritional data set

The statistical analysis of ratio-based indices (ADC and MEC) is common in studies of the digestive physiology of vertebrates; however, recent criticism has questioned the reliability of this technique. We reanalysed a nutritional data set using ANCOVA, and compared the results to a previous analysis performed on ratio-based indices. Our comparison suggested that the analysis of ratio-based indices failed to identify some significant effects, and identified significant spurious relationships that led to erroneous biological interpretations. We used bicoordinate 'utilization' plots (Raubenheimer & Simpson 1994) to aid in interpreting factors that influenced the efficiency of digestion, and we discuss some advantages and limitations of this approach. Predictive relationships based on regression analyses of ratio-based indices led to errors in the estimation of assimilated energy from 1% to 8%, and errors in the estimation of utilized energy from 3% to 20 times 4%. Our reanalysis suggests that errors due to the statistical analysis of ratio-based indices may be more serious than is generally appreciated.