Paleonet: Foram Statistics

Ken Finger kfinger at
Thu Oct 11 18:06:17 UTC 2012

Hi Marty,

Your comprehension of, and comfort with, statistical applications is way beyond mine, so I am compelled to seek out your advice. Now that I'm no longer the JFR editor, I've found the time to return to my research that took a 3-yr hiatus. I am trying to correlate 28 Neogene benthic foram assemblages from "spot samples" taken in several different areas over a latitudinal span of 1000 km. These are from several geographically separated geologic units thought to be of similar age (some think the units may be synonymous), but stratigraphic control is minimal and uncertain because the outcrops consist of horizontal beds and planktic markers are scarce. And there is a lot of disagreement regarding the age(s). The assemblages have many species in common, but there is little consistency among which specimens dominate (common and abundant, >25%). I've used PAST software to run a few cluster analyses on presence/absence matrices where I halved their size by first excluding the very rare species (those limited to no more than 1 specimen in any assemblage), then excluding also the rare species (those limited to <1% of each assemblage), resulting in matrices of 176 and 173 species, respectively.  I've "blindly" applied the similarity coefficients of Jaccard, Shannon-Weaver, and Simpson, as well as Ward's method, but each resulted in an assortment of first- and second-tier groupings in which only 16 of the assemblages paired matched nearby localities or geologic units. I suspect the geographic mismatching of the other half is related to the bathymetric mixing that is apparent in all of the assemblages, or possibly age differences. Might you know if there is a particular mode of cluster analysis or other multivariate method is more appropriate for trying to make sense of my data? Any comments/suggestions would be most appreciated!


More information about the Paleonet mailing list