QUB | Archaeology and Palaeoecology | The 14Chrono Centre
2 Zonation (off) [off] Enter '0' for on, or '1' for offThis section carries out zonation of pollen stratigraphic data, by splitting using binary or optimal techniques (each by minimizing sum-of-squares or information content criteria for making the splits), or by agglomeration using constrained cluster analysis, based on squared Euclidian dissimilarity (CONISS), developed by Grimm (1987), or on information content dissimilarity (CONIIC).
The binary approach splits the data set into successively small groups by splitting existing zones. Results for any given number of zones are thus an extension of results for all results with fewer zones. The optimal approach starts afresh for each successive number of splits, and thus there will not necessarily be any correspondence between results for division into diferent numbers of zones. The principles are discussed fully in Birks & Gordon (1985). Optimal splitting can be very time-consuming. However, in principle and in practice it is more satisfactory than binary splitting, and I now use it routinely (Bennett et al. 1992).
CONISS and CONIIC are based on cluster analysis, with the constraint that clusters must be based on agglomeration of stratigraphically adjacent samples. See Grimm (1987) for details of CONISS. CONIIC differs in using the information content statistic (see Prentice (1980) instead of sqaured Euclidian distance to form the initial dissimilarity matrix.
Results, as defined zones, are written to the ZONE file, unless the data have been shuffled (see menu Mc.6), and will be used and plotted on the current diagram. If the data are not being shuffled and a ZONE file is not available, zonation will not be carried out. A data analysis file may be set up (menu Ec) and, if so, will include details of the numerical side of the zonation. Results from CONISS are laid out more or less as in Grimm (1987, Appendix), except that depth labels are used instead of markers for depths. No judgement is made about subzones: that remains a subsequent matter for the analyst (Grimm 1987).
For CONISS and CONIIC, a dendrogram showing the formation of the clusters, is drawn at the end of the plot, after the final column of zones. If there is insufficient room on the final page, it will be plotted on a new page. It can be moved and (re-)combined with other plot components using pscomb.
It would be usual to apply a threshold value (menu Mi), and recalculate the dataset for analysis to proportions of the sum of types included (menu Mj).
The output tables mark, with `*', zonations with variances considered to be interpretable by exceeding values generated by a broken-stick model of the distribution of variance amongst the various components, as has been done with the eigenvalues of a PCA (Legendre & Legendre 1983; Jackson 1993). The idea is that each successive `split' accounts for part of the total variance in the dataset, and this part is tested against the portion expected from a broken-stick model. Strictly, I suspect that this may not be applicable to optimal splitting, since each level of zonation is begun anew, and is not part of a previous level of zonation. The zonations marked may thus be considered to have some statistical validity. See Bennett (1996) for more details.
If 0 is selected, you will see the following submenu:
1 Number of zones wanted [10] 2 Method for obtaining zones [1] 3 Name for output file with zone details [a:\dallZONE] 4 Prefix for zone labels [D-] 5 Data transformation [0] 6 Use a randomized dataset (off) [off] 7 Modelling of zonation (off) [off] 9 Leave this menu Q Return to main menuFollow the links shown for more details.
Return to main menu.
Archaeology and Palaeoecology | 42 Fitzwilliam St | Belfast BT9 6AX | Northern Ireland | tel +44 28 90 97 5136