Methods for correlation analysis of toxins and species present in random samples of phytoplankton.

We developed and tested two new procedures to find which species present in samples of phytoplankton is responsible for the production of a toxin. The procedures represent a different form of correlation analysis that uses information on the presence or absence of the toxin, and on the relative abundance of each species of plankton in the samples. The efficiency of the algorithms is tested by random process simulation. The algorithms were clearly superior to known techniques dealing with correlations between binary variables to show the toxin producing microorganism. We used experimental toxin isolation from phytoplankton as an example of practical success using the most efficient of the algorithms tested.