Testing positional behavior of malagasy lemurs: a randomization approach.

Studies of primate positional behavior have only occasionally employed statistical testing because these data sets violate the assumptions underlying most standard statistical tests. As currently practiced, both common methods of collecting data on positional behavior involve sequential observations of a limited number of individuals, leading to data sets consisting of interdependent observations. Data collected from sequential observations may not be as dependent as currently accepted; however, approaches are available to deal with such data. Some previously applied tests are discussed, and an alternative strategy, a randomization test, is demonstrated. Randomization tests are based on observed rather than theoretical distributions. Thus, they are robust when applied to data sets which do not adhere to the strict assumptions inherent in standard tests. Randomization procedures suggested by Manly ([1991] Randomization and Monte Carlo Methods in Biology [New York: Chapman and Hall]) and Edgington ([1987] Randomization Tests [New York: Marcel Dekker, Inc.]) are applied to data collected for three species of lemurs studied at Ranomafana National Park, Madagascar (Lemur fulvus, Lemur rubriventer, and Lemur variegatus). Standard chi-square tests indicate statistically significant differences between these species in nearly all aspects of locomotor, postural, and substrate use behavior. In contrast, the randomization tests showed a significant difference only in the postures used by these lemurs. Lemur variegatus uses hindlimb suspensory postures much more often than L. fulvus or L. rubriventer, and the latter species use vertical clinging more often than L. variegatus. The results from the randomization tests are considered more reliable because by using individuals rather than observations as the units compared, they limit the problem of interdependence of observations and thus provide a conservative test. The particular randomization tests used also explicitly test the proposition that individuals of a species are more similar to each other than to individuals of other species. The lack of significant results is not surprising in light of the small number of individuals studied and the great deal of variability within a species for most of the variables considered.