Accuracy and power in the statistical analysis of fluctuating asymmetry: effects of between-individual heterogeneity in developmental instability

A non-exhaustive literature search revealed that samples often show heterog eneity in the underlying developmental instability. As a consequence, the d istribution of the signed asymmetry is leptocurtic. Simulations presented i n this paper showed that a recently developed method to test heterogeneity in FA (likelihood ratio test of REML mixed regression models) has inflated type I error rates, whereas Levene's test suffered power reduction. The lat ter appeared to be the result of a lower accuracy of the estimation of popu lation level FA. These effects became stronger with increasing leptokurtisi s. In contrast, the estimation of R became more accurate and precise with i ncreasing heterogeneity (and thus expected value of R). The estimation of s mall values of R is subject to extremely large sampling variation and were biased towards lower values. Implications for the analysis of FA at the ind ividual and population level are discussed.