An assessment of linear discriminant function analysis as a method of interpreting fossil pollen assemblages

4Citations
Citations of this article
8Readers
Mendeley users who have this article in their library.

This article is free to access.

Abstract

Linear discriminant function analysis (LDF) is assessed as a method of interpreting fossil pollen assemblages from Mt Hauhungatahi, Tongariro National Park, and Ohakune-Horopito, central North Island, New Zealand. Pollen types selected by stepwise discriminant analysis were used in LDF to predict the vegetation type (forest, dense scrub, open fernland, or open tussockland) and canopy type (closed or open) represented by fossil pollen assemblages from an array of sites. Chi-square Goodness-of-fit tests were used to test prediction made by LDF against the vegetation type suggested by the common “standard” method of assessment of the fossil pollen spectra, which is carried out by visual inspection of stratigraphic pollen diagrams. Some sites showed highly significant differences, with many samples obviously misclassified by LDF, especially those from montane forest vegetation. This misclassification occurred mainly because the montane forest samples were so different from the remainder that they were excluded from the prediction set, so that the pollen types most likely to be predictors were not employed. Overall, however, more sites showed no significant differences. The LDF method is confirmatory rather than providing a clear improvement, and its validity will depend strongly on the correct choice of discriminators. © 2003 Taylor & Francis Group, LLC.

Cite

CITATION STYLE

APA

Horrocks, M., & Ogden, J. (2003). An assessment of linear discriminant function analysis as a method of interpreting fossil pollen assemblages. New Zealand Journal of Botany, 41(2), 293–299. https://doi.org/10.1080/0028825X.2003.9512848

Register to see more suggestions

Mendeley helps you to discover research relevant for your work.

Already have an account?

Save time finding and organizing research with Mendeley

Sign up for free