QUB | Archaeology and Palaeoecology | The 14Chrono Centre
0 Rate-of-change analysis (off) [off] Enter '0' for on, or '1' for offRate-of-change analysis was introduced by Jacobson et al. (1987) and Jacobson and Grimm (1988), and further explored by Bennett & Humphry (1995). It involves measuring the dissimilarity between adjacent pairs of samples, and then relating that to the temporal difference between the samples. The methods of Jacobson et al. (1987) and Jacobson and Grimm (1988) are not well-explained, but it appears that they first smoothed their pollen data, then interpolated to give pollen spectra at even intervals of time along the sequence, then measured dissimilarities. The method used here is to measure the dissimilarity between adjacent samples, and divide this measure by the temporal difference between the samples (Bennett et al. 1992). This is simpler, and involves much less tinkering with the data. However, dissimilarity measures are non-linear (in a sequence of three samples, the dissimilarity between samples 1 and 3 will not be the same as the sum of the dissimilarities between 1 and 2 and between 2 and 3). This should not be a serious problem where the time difference between samples is approximately constant, or changing smoothly and slowly.
The dataset can be smoothed before doing the analysis. The number of points smoothed can be 3, 5, 7, or 9. Smoothing is carried out by a least-squares line-fit, with 2-terms (straight line) for 3 points, or 4-terms (cubic polynomial) for the others.
Results are available for immediate plotting, and will be included with the rest of the dataset for saving to the data analysis file (menu J). The `taxon name' given is name of the dissimilarity measure used (see below). Depths must be either presented as ages (see Data preparation) or converted to ages (menu La) in order to obtain rates of change. This means that on the psimpoll run that does the calculations, the output will have to be plotted against age.
If 0 is selected, the following display appears:
1 Dissimilarity measure (1) [1] 2 Smoothing parameter (0) [0] 3 Modelling of rate of change (off) [off] 9 Leave this menu Q Return to main menu
1 Chord distance
2 Information statistic
3 Chi-squared coefficient 1
4 Chi-squared coefficient 2
5 Manhattan metric
6 Euclidian distance
7 Angular separation
8 Standardized Euclidian distance
Enter number <Enter>
There are many dissimilarity measures
(Prentice 1980),
but currently I have implemented just these eight after experiments
by Roger Humphry
(Bennett & Humphry 1995).
Choice 1 is the classic measure, and the one used by
Jacobson et al. (1987),
Jacobson & Grimm (1988), and
Bennett et al. (1992). Try them all.
Return to main menu.
Archaeology and Palaeoecology | 42 Fitzwilliam St | Belfast BT9 6AX | Northern Ireland | tel +44 28 90 97 5136