Exploratory path analysis with applications in ecology and evolution.

In this article, I first describe some recent developments in the identification of the structure of dependencies among variables in multivariate data relevant to exploratory path analysis. I then introduce a bootstrap modification of one important method (the SGS algorithm) that is designed to improve error rates of exploratory path analysis in the small data sets that are typical of studies in ecology and evolution. Monte Carlo results indicate that this modified technique can find path models that are close to the true model even in very small data sets. The bootstrapped SGS algorithm is then applied to a previously published data set involving attributes affecting seed dispersal in St. Lucie's cherry.