Corr to target members - Matching members of two collections to each other

Last update: May 2 2019

Corr to target members - How members of one collection match best towards members of another collection

The "Corr to target members" matrix was a heavily desired tool during a crossdating project we run in 2016 with more than 500 samples originating from a wide area of north-western Europe. Samples could indeed be crossdated but often towards several curves and we wanted a way to quickly find the best matches and also to easily document our findings.

Note: Target members in leftmost column. The table may be created with a big number of columns and of rows!

This matrix shows how your collection members match towards the members of a synchronized target collection containing typically several reference curves. The match is done either to check the "current offset" (current dating) or to find ”the best matching offsets". The result is a matrix with your reference curves on the rows and your samples in the columns. This tool is particularly useful when you have many samples of unknown origin, when wood has been imported from a wide geographical area. The matrix can then be used with a high number of columns (undated samples) combined with sorting the best matching (chocolate colored) samples towards the left part of the column to quickly find easy crossdatings. The matrix can also be used to document the sample matchings for a single site.

Below each column in the matrix, you will also find a small table with one or more rows showing the "most popular" dating years and their crossdating quality (T-values), for example

Count MaxT MeanT Year 
    2  4.6   3.7 1980 
    4  9.0   6.5 1978T 
Meaning that four of our reference samples had their best crossdating towards this sample at the same endyear 1978 with a mean T-value of 6.5 and a maximum of 9.0 This type of small table has also been added to the "With condensed report" of the "(Test towards)..undated ref." command.
Count MaxT MeanT Year 
    2  4.0   4.0 1019 
    2  4.1   3.5 1740 
    6  4.5   3.6 1990 
    2  5.1   4.9 1843 
    2  9.9   9.1 1821
Earlier versions of CDendro only had the year and the count column. Now the MeanT and MaxT reveals that the correct dating is at 1821 though it has only a count of 2.

How to get matching references into a new collection?

When both collections contain many members it might be difficult to interpret the results. E.g. with a "COUNT-report" like shown above for one of your samples,you might say "Which target members said that 1978 makes a good match? - Give me a collection with them all!" A way to achieve this is to open your sample in a separate window (double click on the member) and then date your sample to 1978. Then from the (synchronized) target collection you run the command "Test towards ref." That will show how this sample - when dated to 1978 - will match towards all of your (synchronized and dated) target collection members. By sorting the T-column you can focus on the best matching members and have only them checked so you can save them in a separate new collection.

Please note that your target collection does not have to be correctly dated, it might very well have a floating date (EndYear) like 8000. But it has to be synchronized!



Copyright © 2019, Cybis Elektronik & Data AB, www.cybis.se