The Development version of CDendro

You may post any question related to CDendro in this meeting.
Post Reply
Lars-Ake
Site Admin
Posts: 87
Joined: 21 Dec 2008, 23:24
Real name: Lars-Ã…ke Larsson
Location: Saltsjobaden (Stockholm), Sweden
Contact:

The Development version of CDendro

Post by Lars-Ake »

There is now a new development version of CDendro available for testing.

The "Condensed report" of the button-command "Test towards reference" (in a collection) has been somewhat modified:
When running towards an UNDATED reference, there is now a small table printed at the end of the condensed report, e.g.

Code: Select all

Year numbers occurring more than once:
 1792   11
 1769    3
 1824    2
 1921    2
 2010    2
When running towards a DATED reference, the dating information of the member can now either
a) be removed or
b) be shown in a column as before or
c) given as an interval which may either be placed in a column or
d) at the end of the report line (as in the example below).

Code: Select all

                    T-  Over
           Corr   Test   lap
all...     0.29   3.31   121   based on 37 members
3EP281AV   0.50   6.23   121   Estonia Est002 (EXTRATERR) (1516-1998)
BratPIr    0.43   4.91   107   Mälardalen (Bråthen), questionable segments removed (1153-1904)
ANEBYALL   0.35   2.57    48   Aneby in Småland (1745-2003)
The "With condensed report" checkbox of a collection has a Tooltip about these options.
Note: The development version of CDendro is available for testing by CDendro customers who have either upgraded to or recently bought CDendro version 7.
Lars-Ake
Site Admin
Posts: 87
Joined: 21 Dec 2008, 23:24
Real name: Lars-Ã…ke Larsson
Location: Saltsjobaden (Stockholm), Sweden
Contact:

Re: The Development version of CDendro

Post by Lars-Ake »

There is now a new development version of CDendro together with a new version of CooRecorder.
Main feature to traverse a crack with CooRecorder is reported in another posting.

In CDendro the Correlation coefficient (and TTest-values) can now be shown like this:
Image
It may be useful to convince a customer about the reliability of a dating.
(It may be noted that negative corr-coeff values can be shown as an option - though it only makes the picture look scribbled.)

The lattitude/longitude field of a .rwl file, as set in a collection, may now extend to 11 characters to allow a longitude > 99 degrees. (Only 10 characters specified in ITRDB format specification!?)
Lars-Ake
Site Admin
Posts: 87
Joined: 21 Dec 2008, 23:24
Real name: Lars-Ã…ke Larsson
Location: Saltsjobaden (Stockholm), Sweden
Contact:

The Development version of CDendro

Post by Lars-Ake »

The development version of CDendro has been updated.
Lars-Ake
Site Admin
Posts: 87
Joined: 21 Dec 2008, 23:24
Real name: Lars-Ã…ke Larsson
Location: Saltsjobaden (Stockholm), Sweden
Contact:

The Development version of CDendro

Post by Lars-Ake »

There is now a new CDendro Development version available for testing. It is dated March 19 2009.

The main feature is the Besancon index E algorithm now available as one of the main normalization algorithms used for plotting, like Proportion of last two years growth, Hollstein and Baillie/Pilcher. The Besancon index can also be selected for display in the "Sort best matches" table. The Besancon algorithm is described in the big textbox at Settings/Options for normalization and matching

The article "What is a good TTest value?" has been updated with data related to the Besancon normalization.

I have tried to demonstrate for myself that one of these normalization methods is in some way better than the others. I've failed!
It seems to be a matter of personal preferences which algorithm to use for the curves and crossdating. These methods all behave a little bit differently but on the average they all seem to be of the same quality.

For daily work with crossdating, I have all the "Sort best matches" checkboxes checked for display in the Options for normalization.../Sort best matches frame as shown below.

Image

As TTest values are only available for that method used for the curves display, there is now a TTest Calculater available in the Help menu.

Image

The calculator may sometimes be useful when you want to "translate" a single correlation coefficient value into a TTest value.
Note that the calculator is floating on the top of your screen, so you may use it also when looking at e.g. a crossdating table in your browser.
The calculation is done when you click on the Enter key in one of the fields or when you click the Calculate button.
Lars-Ake
Site Admin
Posts: 87
Joined: 21 Dec 2008, 23:24
Real name: Lars-Ã…ke Larsson
Location: Saltsjobaden (Stockholm), Sweden
Contact:

Re: The Development version of CDendro

Post by Lars-Ake »

There is now still another new development version of CDendro - this time named CDendro 7.2 - available for download and testing.

Various detrending and normalization methods have been added to the program during a long time. I have now rewritten these methods so they are all handled in the same way. This makes program development easier and it also makes it easier to compare various crossdating methods more systematically.

A new detrending method based on a polynom algorithm of selectable order has been added. From time to time I have got requests for such a mechanism.
Now that I can compare that method with that old "Heavy detrend" method of CDendro, I can see that they are both of about the same quality...
"Heavy detrend" divides each ring width value with the mean value of the nearest 25 (now selectable) ring widths. With the new 7.2 version you can easily see that the results will be about the same:
Image
Selected polynom orders can be automatically saved as an encoded comment within "a parent collection" with samples.
This may be of interest if you try to carefully build a mean value ring width file.

For a more rough method when building a mean ring width curve out of a big collection, I recommend the NegExp detrending method for collections in CDendro.

(The curves above are based on a PCAB (spruce) tree measured by Torbjörn Axelson.)
Lars-Ake
Site Admin
Posts: 87
Joined: 21 Dec 2008, 23:24
Real name: Lars-Ã…ke Larsson
Location: Saltsjobaden (Stockholm), Sweden
Contact:

Re: The Development version of CDendro

Post by Lars-Ake »

There is now a new development version of CDendro available for customers who want to test it.
/Lars-Åke
Post Reply

Who is online

Users browsing this forum: No registered users and 27 guests