KELLY, MARTIN G. Department of Biology, Buffalo State College, Buffalo, NY 14222-1095. - Comparing non-parametric and parametric methods of analysis for natural selection.
Natural selection occurs when nature selects the fittest individuals
based on their ability to survive and reproduce. Natural selection is
the major force for change in the gene frequencies of a population
that leads to evolution. The goal of this research is to compare
parametric and non-parametric methods of analyzing phenotypic
selection. By comparing the published results derived by the
parametric method of analysis (Arnold and Wade, 1984; Lande and
Arnold, 1983) with the non-parametric analysis of the same data
(Schluter, 1988), the results derived from both methods of analysis
can be compared to see if they yield similar (or different)
descriptions of natural selection. Because the parametric method
requires a quadratic model to describe the quantitative relationship
between phenotype and fitness, it can only be used to describe natural
selection for data appropriate to this model. In contrast, the
non-parametric method of analysis is a regression method that does not
require a prior model for the quantitative relationship between
aspects of an organism’s behavior or appearance and its survival or
reproduction. This research will be done using published data where
natural selection was analyzed using the parametric method of
selection analysis developed by Stevan J. Arnold, Michael J. Wade, and
Russell Lande (Arnold and Wade, 1984; Lande and Arnold, 1983). These
summary data will be reanalyzed with the non-parametric method of
analysis developed by Dolph Schluter (Schluter, 1988). I will
summarize the results from the two methods of analysis to determine if
similar (or different) functions describing natural selection were
produced by two independent methods of analysis.
Key words: non-parametric, parametric, phenotypic selection, univariate analysis