Statistical methods for the analysis of floral scent data

We use the CNESS (chord-normalized expected species shared) distance index, ranging between 0 and the square root of 2, to determine the differences between the single samples. CNESS is a metric version of Grassle and Smith's (1976) NESS similarity index, which was originally built to compare faunal or floral samples. NESS and CNESS can be regarded as more generalized forms of the Morisita index (Morisita, 1959). Wolda (1981) investigated many quantitative similarity indices and found that all but one, the Morisita index, were strongly influenced by sample size and diversity. The only disadvantage of the Morisita index is the high sensitivity to changes in the abundance of the most abundant species (Wolda, 1981), or in our case, the most abundant compounds. However, this problem can be solved by using NESS or CNESS, which can be adjusted, by altering the m-parameter to emphasize the importance of minor compounds in the data (Wolda, 1983).

We calculate CNESS indices using the updated version of the COMPAH (Combinatorial Polythetic Agglomerative Hierarchical Clustering) program (Boesch, 1977), provided by Gallagher at UMASS/Boston (http://www.es.umb.edu/edgwebp.htm). We determine the best CNESS-m using the method described in Trueblood et al. (1994). We use nonmetric multidimensional scaling (NMDS) in the STATISTICA 5.1 package to detect meaningful underlying dimensions and to visualize similarities between samples (see Borg and Lingoes, 1987). To evaluate how well (or poorly) the particular configuration produces the observed distance matrix the stress value is given. The smaller the stress value, the better is the fit of the reproduced distance matrix to the observed distance matrix (Clarke, 1993).

We use the Mann-Whitney U test to compare intraspecific variation in the chemical profiles to interspecific variation. We therefore compare the pairwise dissimilarities (CNESS, explained above) between individual samples within a species/genus to those between samples among species/genera. A variance component analysis is used to estimate the contribution of single compounds to the obtained total variation (presence/absence and/or amount) between the taxa.

BOESCH, D. F. 1977. Application of numerical classification in ecological investigations of water pollution. Report prepared for US EPA Ecological Research Series (EPA-600/3-77-033). Available as PB269 604 from: National Technical Information Service. U.S. Dept. of Commerce, Springfield Virginia, USA.

BORG, I., AND J. LINGOES. 1987. Multidimensional similarity structure analysis. Springer, New York, New York, USA.

CLARKE, K. R. 1993. Non-parametric multivariate analyses of changes in community structure. Australian Journal of Ecology 18: 117-143.

GRASSLE, J. F., AND W. SMITH. 1976. A similarity measure sensitive to the contribution of rare species and its use in investigation of variation in marine benthic communities. Oecologia 25: 13-25.

MORISITA, M. 1959. Measuring of the dispersion of individuals and analysis of the distributional patterns. Memoirs of the Faculty of Science Kochi University, Series E 2: 215-235.

TRUEBLOOD, D. D., E. D. GALLAGHER, AND D. M. GOULD. 1994. Three stages of seasonal succession on the Savin Hill Cove mudflat, Boston Harbor. Limnology and Oceanography 39: 1440-1454.

WOLDA, H. 1981. Similarity indices, sample size and diversity. Oecologia 50: 296-302. WOLDA, H. 1983. Diversity, diversity indices and tropical cockroaches. Oecologia 58: 290-298.

WOLDA, H. 1983. Diversity, diversity indices and tropical cockroaches. Oecologia 58: 290-298.