How do violations of model assumptions influence diet estimates?

What is it about?

Understanding animal diets is a critical aspect of ecology and quantitative fatty acid signature analysis (QFASA) has become an important method of estimating predator diets, especially for marine species. QFASA involves two primary assumptions: (1) prey species potentially eaten by a predator are included in the prey data and (2) “calibration coefficients” used to adjust for a predator’s differential metabolism of fatty acids are known perfectly. The first assumption can be satisfied by obtaining additional prey samples, if necessary, but the validity of the second assumption is impossible to verify. We investigated the accuracy of QFASA estimators when these two assumptions are violated. Our primary finding is that estimators based on the Aitchison distance measure are less sensitive to errors in the calibration coefficients than estimators based on the Kullback-Leibler distance measure when assumption violations are not severe. Consequently, use of the Aitchison distance measure is recommended because it offers more protection against the assumption whose validity cannot be verified.

Featured Image

Why is it important?

Our research provides new information about how QFASA diet estimators perform when model assumptions are violated. These findings were used to develop recommendations to proactively minimize potential bias in diet estimates that will be valuable in future studies of animal diets based on QFASA.