Quantifying fluctuating asymmetry: Not all methods are equivalent

Academic Article

Abstract

  • Although numerous methods exist for quantifying fluctuating asymmetry, they fall into two categories: (i) statistics employing the difference between the counts or measurements on the left and right sides (L - R or R - L), and (ii) use of the correlation coefficient between the counts or measurements for the two sides. Estimates of asymmetry obtained using various treatments of the difference between sides are quite concordant in quantifying the relative asymmetry of different groups. The correlation coefficient, on the other hand, may produce a quantitative estimate of asymmetry which is not at all concordant with the other estimates of asymmetry. This is because the strength of the correlation between bilateral characters and thus the magnitude of the correlation coefficient, r, depends on the variability characteristics of the data with respect to the line of perfect symmetry (leading diagonal) in a plot of the data. Scatter of points away from the diagonal reduces the magnitude of r and is indicative of asymmetry. Variability along the axis of the leading diagonal also affects the magnitude of r but does not represent asymmetry. Both hypothetical and actual data are presented which demonstrate that the correlation coefficient is highly sensitive to character variability whereas the other estimates of asymmetry are not.
  • Authors

    Published In

  • Growth  Journal
  • Author List

  • Angus RA
  • Start Page

  • 337
  • End Page

  • 342
  • Volume

  • 46
  • Issue

  • 4