Researchers have suggested fluctuating asymmetry (FA) as an indicator of environmental stress and have usually tested this assertion by examining relations between FA of single traits and stress. Fluctuating asymmetry stress relations are real but are typically weak and difficult to detect. Researchers would like to maximize the probability of detecting FA-stress relations when they exist. We assert that analyses based on the FA of multiple traits may provide better methods for detecting stress. In this article, we used computer simulations to compare the ability of six analyses to detect differences in FA between stressed and unstressed populations. We show that the optimal analysis depends upon the underlying form of the FA distributions. We also show that two of the analyses had inflated Type I errors in some situations. Finally, we quantify the advantage of our preferred analysis over those of single-trait FA in detecting stress.

No abstract

Parasites may be expected to become locally adapted to their hosts. However, while many empirical studies have demonstrated local parasite adaptation, others have failed to demonstrate it, or have shown local parasite maladaptation. Researchers have suggested that gene flow can swamp local parasite-host dynamics and produce local adaptation only at certain geographical scales; others have argued that evolutionary lags can account for both null and maladaptive results. In this paper, we use item response theory (IRT) to test whether host range influences the likelihood of parasites locally adapting to their hosts. We collated 32 independent experiments testing for local adaptation, where parasites could be assigned as having either broad or narrow host ranges (BHR and NHR, respectively). Twenty-five tests based on BHR parasites had a significantly lower average effect size than seven NHR tests, indicating that studies based on BHR parasites are less likely to demonstrate local parasite adaptation. We argue that this may relate to evolutionary lags during diffuse coevolution of BHR parasites with their hosts, rather than differences in experimental approaches or other confounds between BHR and NHR studies.

No abstract

Variable reporting of results can influence quantitative reviews by limiting the number of studies for analysis, and thereby influencing both the type of analysis and the scope of the review. We performed a Monte Carlo simulation to determine statistical errors for three meta-analytical approaches and related how such errors were affected by numbers of constituent studies. HedgesÕ d and effect sizes based on item response theory (IRT) had similarly improved error rates with increasing numbers of studies when there was no true effect, but IRT was conservative when there was a true effect. Log response ratio had low precision for detecting null effects as a result of overestimation of effect sizes, but high ability to detect true effects, largely irrespective of number of studies. Traditional meta-analysis based on HedgesÕ d are preferred; however, quantitative reviews should use various methods in concert to improve representation and inferences from summaries of published data.