In an ecological context, concordance may be defined as the tendency for paired values of some parameter, such as the annual productivity of bird species, to show similar directions and magnitudes of deviation from the mean. Where concordance among populations is high, there is an implied similarity of the ecological factors affecting performance. Conversely, if populations behave discordantly, dissimilarity of underlying ecological factors is likely. In evaluating birds as indicators of the marine environment, the biologist typically is confronted with a three-dimensional array of observations (species, areas, and years) in which there are more missing values than filled cells. This frustrates attempts to analyze concordance using existing methods (e.g., Kendall's coefficient, or correlation combined with cluster analysis), which are either impossible to apply to incomplete data sets or potentially misleading when applied to incomplete data sets. I suggest an alternative method for analyzing concordance that makes maximal use of available data. For a given data set partitioned into the smallest units containing information about concordance, one computes an index of concordance using a regression approach and tests for significance using randomization methods. This procedure would seem to have wide application to ecological studies generally and to seabird monitoring in particular.
