Various tips in using CDendro and CooRecorder

Last update: Oct 29 2013.

Contents

Quick switching between the CooRecorder and CDendro programs
TTest/T-value calculator
SECONDARY reference

The Workbench tab

Find best match within a limited region of time
Standard deviation for blockwise correlation coefficients along a curve match
Blockwise correlations and Sorted SetsSampleTo
Options for matching and normalization - Alternative crossdating methods
New normalization algorithm based on mean values of the best sliding frames algorithms: P2Yrs, Baillie/Pilcher, Cross84, Besancon IE.
Which column controls the sort order for "Workbench/Make whole sample correlation analysis"?

The Display curves tab:

Standard deviations
Zero rings
Any ring (not only a zero ring) can now be removed in CDendro
Translate diagram data to coordinate data - find the point to add or remove
Checking the correlation for a segment of the curve. Cutting off part of a curve/Trimming

Collection handling

Opening a group of collections
Sorting order for collection members


Sort collection members by correlation coefficient or T-value
Sorting checked and unchecked members separately
Delete selected members
Collection Comment handling
Remake a multi-radii multi-sample collection into a multi-sample collection
Time lines diagram
Collections/Create list of members for all open collections" produces a detailed report of how your collection members are dated.
More than one mean-value-sample from the same collection
Crop of a collection
Collections - Changing the identity of a member sample
BULK addition
Plotting collection members in parallel
Copy a file header from one collection into another
Invert member checking in a collection
.rwl file precision
The cross correlation matrix as a chocolate graphic
Using a .rwl or .fh file vs using a .fil file when saving a collection
When saving a collection as .fil, also unchecked members can be included together with their checking status
Statistical reports

A few more old tips

Exporting ring width data from Excel
A tip on handling file name extensions when saving data from Internet Explorer
Printing Workbench correlation reports and reports on collection properties
On reading too small text

  • Quick switching between the CooRecorder and CDendro programs:
There are mechanisms in CDendro and CooRecorder to easily switch between these two programs. The current .pos file will be opened in the other program.
TTest/T-value calculator
A small TTest calculator is available under the Help menu command. This will float above all Windows applications so you can use it also in other programs.
SECONDARY reference
The normalized curve of a SECONDARY reference can be shown. This is useful when you sometimes want to look at three matching curves at the same time. To set a sample as the secondary reference, see first that its offset is correct relative to the main reference. Then right-click on the Select as reference button below the curve diagram. (The button has a tooltip about this.)
To release the secondary reference, use the menu command Samples/Set no Secondary reference selected.

The Workbench tab:

Find best match within a limited region of time
You may limit the time interval to be searched for matches. Right click on the button Make whole sample correlation analysis. The search is limited by a First year and an End year. You can use the upper half of the "Set years to limit search"-window to calculate suitable First and End  years.
Standard deviation for blockwise correlation coefficients along a curve match
The standard deviation for blockwise calculated correlation coefficients along two matching curves can be displayed. Though you have to enable an option to make these values shown.

A low correlation coefficient value combined with a high standard deviation value may indicate a missing or false ring. It may also be a useful measure when looking for matches in distant chronologies when the dating is unknown. A low standard deviation value combined with a reasonable but anyhow low correlation coefficient value may then indicate an overall acceptable match.

The correlation coefficient is a measure of how well two curves match at a certain overlay position. The main measure is then the coefficient calculated for the full overlapping length of the curves at the given position. When you look at a match segment by segment (block by block), the coefficient may have quite different values along the curves. The standard deviation of this sequence of correlation coefficient values may be informative when you inspect a list of possible low quality matches. The length of the blocks and the distance between them can be set. Please note that the blocks may overlap, i.e. the correlation coefficient be e.g. calculated every 10th year but then calculated for the last 30 years. A shorter block length will give you a higher standard deviation.

Blockwise correlations and Sorted SetsSampleTo
Block correlation analysis result "SetsSampleTo" sorted in a separate list at the end of the report.

There is an option "Extended block length test" of the "Make block correlation analysis" of the Workbench tab.
When a block has been tested, it is successively lengthened with the blockdistance and then again with the blockdistance. These new blocks are tested for best and next best position within the reference. Result is presented as a sorted "sets sample to year number-list". Missed or erroneous rings may be discovered with this tool.

Options for matching and normalization - Alternative crossdating methods

You can select which crossdating method to use for sorting out the best match.
You can also have data from several methods shown in the crossdating table.

Note: In cases when either the sample or the reference is created directly with the "Create Mean value" button in a collection, then the data used for the different columns above are based on differently calculated values. Then there are actually two types of values available from the collection: Mean value normalized data (based on the normalization method then used) and Mean value ring width data. I.e. already calculated normalized data is only available for the normalization method then used. Normalized data to use for correlation calculus for also the other methods (columns in the table) is then calculated from the ring width mean values. This means that you might get tables with slightly different correlation coefficient values depending on which normalization method you have used for sorting.

New normalization algorithm based on mean values of the best sliding frames algorithms: P2Yrs, Baillie/Pilcher, Cross84, Besancon IE.

When crossdating pine and spruce in Scandinavia we have found that the P2Yrs method is usually the best to use. Though for some continental oak wood we have found the other standard methods to be very relevant and of considerable help. Accordingly I have added a new normalization method to CDendro: "Mean of sliding frame algorithms" with correlation coefficient values based on a mean of calculations for the P2Yrs, Baillie/Pilcher, Cross84 and Besancon Index E methods. When this method is selected, the normalization curves (for visual inspection) are based on the P2Yrs method, though all correlation coefficients (also blockwise) are calculated as mean values of all the four methods. See also the section Methodology/Fooled by narrow rings and the normalization method

Note: With several methods checked in the right "Check to display"-column as shown in the picture above, CDendro will present you with correlation coefficients and TT-values for all your checked methods when you click the "Make a whole sample analysis" button on the Workbench, see the previous picture above with the heading "Find best match within a limited region of time".

Which column controls the sort order for "Workbench/Make whole sample correlation analysis"?

When you click the button 'Make whole sample correlation analysis' on your samples Workbench tab a crossdating operation is started within CDendro. Now the position of your sample's ring width curve is tested for all possible positions overlaying the ring width curve of your current reference and a correlation (coefficient) value is calculated for each such position. The best values are then sorted out and presented to you in a table. During this operation not actually the ring widths are used, but instead "normalized values" calculated out of the ring width values. There are several normalization methods to choose among. In the picture above the P2Yrs method is selected. Now a good correlation coefficient value tells only if the match looks good or not. A TTest value is also calculated where the 'overlap' between your curves are taken into accout. An overlap of only 35 years with a correlation coefficient of 0.6 (tails matching each other) is not as good as an overlap of 100 years and a correlation cofficient of 0.55. So normally it is advisable to select your best matching alternatives based on the TTest value. Therefore 'Sort by TTest' is checked above!

There is a '**'-mark above the column used for sorting out the best matches.
There are cases when you nevertheless want to get the best matches sorted out based on the correlation coefficients. By first unchecking 'Sort by TTest' in the 'Options for matching and normalization' window you can get a report as shown above. Note that the '**'-mark is now above the CorrC column! After such an analysis you should restore 'Sort by TTest' to its checked status.

Note: In the 'Options for matching and normalization' window there is a rightmost column 'Check to display' containing checkboxes for each normalization method available. By checking some of these you get extra columns calculated based on these normalization methods. When all columns give you a good correlation value you might trust a proposed dating result. But please note that the sorting is based on only one column! So if you select another normalization method than P2Yrs, you will probably get another sorting sequence between the proposed matches, though the top item (the best match) is probably the same.

The Display curves tab:

Standard deviations
When you click on the button "Create mean value sample" within a collection, both a ring width mean value curve and a normalized mean value curve are created. Also the standard deviations are calculated and displayed when there is no other active reference (which normally is shown together with your sample curve). If there is another reference you can click in the checkbox "ShowStdDev" to see the standard deviation curves laying on each side of both mean value curves.
Zero rings
There are times when you cannot measure the full radius of a sample because a section is missing or
because rings are invisible or ring widths are too thin to be measured.
To insert a zero ring in CDendro, right-click where you want the zero ring inserted.
To remove an existing zero ring, shift-right-click in the curve.
Zero rings inserted or removed in CDendro can be saved in your coordinate (.pos) file and directly reloaded in CooRecorder for e.g. a tighter inspection.
(In the picture above, two zero rings are still missing.)
You can also insert one or several zero rings at a crack or at a diffuse section when measuring in CooRecorder.
You can later adjust the width of such a zero ring gap in the curve diagram of CDendro.
Note 1: Zero ring gaps do not participate in the calculations when finding the best match!

Note 2: Be very careful when reading TTest-values for a sample containing zero rings!
That sample may give you a very high TTest value though it comes from yourself adjusting the curve to match at a certain point!

For more information on zero rings, see the section Handling zero rings

  • Any ring (not only a zero ring) can now be removed in CDendro

Use Shift-Right-Click to remove a ring from a curve!

Insert/Remove zero rings: Since long time there is a mechanism in CDendro to insert a zero-ring (right-click on the curve) and to remove a zero ring (shift-right-click on or near the zero ring on the curve).

Remove a non-zero ring: When there is no zero ring, a shift-right-click will now remove the ring at the point of the click on the curve and open a dialogue window where you can specify HOW to remove the ring. This is shown in the next image. At the top is the curve with that unwanted ring. Next follows how the diagram looks after a shift-right-click on the curve.

A group of radio buttons give you options on how to remove the ring. The curve diagram will be updated correspondingly when you select another one of the radio buttons, so you can directly see which solution gives the best result.

A comment proposal is also given, to allow for documentation of what you have done.

This ring removal mechanism is not available for CooRecorder .pos files. In that case I assume that the user should best update his/her CooRecorder .pos file, save it and use the Reload button (under the curve diagram) to get the update into CDendro.

Translate diagram data to coordinate data - find the point to add or remove

Clicking on a point in a diagram will print coordinates, ring number and year of that point when the origin is a .pos file.
This saves you time when you want to remove or add a point at a certain place in a coordinate file!

There are always situations of "is this a false ring?" when you are in doubt if a border is a yearly full growth season ring or just a ring probably created by early cold weather in late summer followed by a long period of warm weather. When crossdating you have to add or remove (or enable or disable) such points in CooRecorder. With this "diagram-to-coordinate-data-translation" feature you can quickly go from a curve point in the CDendro curve diagram to a measured point in the CooRecorder measurement pane.

An alternative: You might also consider to display your reference curve and your ring width measurement data in CooRecorder! See the section on Using CooRecorder: Under the heading "Dendro measurements:", see the section "Check against a reference curve during measurment!"

Checking the correlation for a segment of the curve. Cutting off part of a curve/Trimming

You may indicate the start (1) and end (2) of a segment of a longer curve and get the correlation value for that segment calculated (3).

That marked-out part of the sample can be copied (4) into a new sample window.
This may be a quick way to remove a mismatching end of a sample.

Writing out a trimmed/cropped .pos file
When Create a sample from block (i.e. copy a segment) is done on a coordinate file, the original coordinate information is also copied.
So a new coordinate file can easily be written out from that copied data (Samples/Save coordinate data as)!
That is fine when you want to keep only a segment of a coordinate file because of bad crossdating.

Collection handling

Opening a group of collections

You can now save the names of your open collections in a special "group-file" and later when you restart CDendro open those files with one single command.

(The same mechanism for saving and restoring of sample windows is much more tricky to implement - so we have to wait for that.)

Sorting order for collection members

You can sort any collection column by clicking repeatedly on its header. First click sorts by increasing order, next click by decreasing order.

Members of collections are by default sorted by their member identity (1) and not on their full file paths.
You can replace an old measument with a new one even if it comes from another directory (2) and you will still keep the correct sorting order to see which members come from the same stem.

When you copy a sample (member) from a collection into the target collection, you will get a warning message if that will overwrite an already existing member in the target collection. This is an easy way to replace an erroneous measurement serie already stored in a big .rwl or fh-file with a new measument serie from a .pos-file.

Note: The member list above is written with quite a big size font. If you prefer to have it with a smaller font size giving you more members visible on the screen, use the menu command Settings/More Settings and uncheck the box "Use big fontsize for collection members".

Sort collection members by correlation coefficient or T-value

  • You may sort a column by clicking in the column head.
Two sortable columns for corr-coeff and TTest values from the last correlation calculus: I often use the "Test towards rest of collection" button to find those collection members that do not match the others so well or "Test towards reference" to check the member matchings towards a reference. By then clicking on the CC column, I get the members sorted in order of lowest or highest correlation coefficient value at the top of the column.

There is an option to hide these two columns (if you do not like them) by checking a checkbox with the Settings/More settings menu command.

Note: You can temporary rearrange your columns by dragging them sideways to where you want them. This is a Windows standard mechanism. For an example see the picture with the timelines diagram below where the EndYear, StartYear and TT columns have been shifted to the left.

Sorting checked and unchecked members separately

By checking the command "Sort checked and unchecked members separately" checked and unchecked members can be separately sorted so that all the checked members are first listed and then the unchecked. You can sort a column by clicking (left-click) on the header of that column, i.e. get all samples sorted by their length by clicking on the "Length"-header.

Note: To get the two-item pop-up menu shown in the picture above, you have to RIGHT-click on one of the column headers.

Delete selected members

The "Delete member" button of a collection has been changed to a "Delete selected members" button:
  • It operates on all selected members of a collection. (i.e. click one member, then hold down the shift key and click another member to select also all members in between. Then click "Delete selected members" to delete (remove) them all from the collection.)
  • This command will now also retain your current member focus, i.e. the "member-lines" to delete are just removed from the screen - all other members keep their relative positions on the screen, i.e. even the current sorting order!

Collection Comment handling

Comment handling has been enhanced so that you can (for checked members) copy e.g. file prenames to the comment field.

This is useful when you add several collections into one big collection but want to retain information of the origin of each member.

Remake a multi-radii multi-sample collection into a multi-sample collection

The collection shown above is a multi-radii, multi-sample collection, i.e. consisting of series from many trees and with one or more series from each tree. The command "Save as new collection with members summed by stem..." will create a new collection with a mean value ring width curve for each tree. I.e. the original multi-radii, multi sample collection is remade into a multi-sample collection.

Time lines diagram

The Time lines diagram with the optional sample depth diagram on top gives you a way to easily demonstrate how the collection members are related to each other. Several of the drawing parameters can be adjusted to make the diagram suited for e.g. a report. There is a button to copy the diagram as a bitmap (raster image) to the clipboard. You can then paste it into MS Word or an image editor.
  • "Collections/Create list of members for all open collections" produces a detailed report of how your collection members are dated.
Collections/Create list of members for all open collections produces a .html report file with a detailed report showing how collection members are dated. The collections can be sorted by EndYear or by name.
For more extensive examples, see the section "Dating reports for published mean value collections:" at our Belfast section

More than one mean-value-sample from the same collection

Create mean value sample of a collection: Right-click to get it in a NEW window!
Right-click on that button will now create a NEW mean value window and also prompt for names of this new and the previous mean value window.
There is a tooltip on the button about right-clicking for this option.
Crop of a collection
New menu command: Collections/Write collection as cropped members to a .fh file. Will cut out the same time segment for all members.
This is a useful tool when a collection is found to be correct within a certain interval but is being corrupt elsewhere.
This tool is also useful if you want to cut out a section of time with a stem density of e.g. at least 5 stems.
Collections - Changing the identity of a member sample

The identity of a member of a collection can easily be changed like the comment or offset.

This means the the members can be easily reorded and the result saved in a .rwl or .fh file.

This is of great importance when samples from the same tree have been given too different identities like 6681 and 6682 instead of 6681A and 6681B.
See the section on Multi-radii collections

BULK addition

BULK addition There is now a command to import all single sample files of a certain type from a directory. We have used that to import all the 9000 separately downloaded Belfast files into one collection. This command was needed because Windows has a limitation meaning that you can only select a maximum of some 250 files in the Open Files dialogue box.

Plotting collection members in parallel

Members of a collection can be plotted as parallell curves.
Vertical dated lines can be set on the diagram - just control-click where you want it.
Control-Click again on a line to remove it!
Both ring width curves and normalized curves can be plotted in this diagram.

Copy a file header from one collection into another

There is a menu command for you to copy the file header from one collection to another.

Invert member checking in a collection

There is a button to invert the member checking of a collection. This is typically useful when you want to split a decadal file into two different collections.
.rwl file precision
.rwl files can be written in 0.001 units, not only in 0.01 units!
This is selected through the menu command Settings/More settings.

The cross correlation matrix as a chocolate graphic

The collection buttons "Find best cc" and "Find current cc" have been replaced by a "Show Cross correlations" button with a dialogue where you can specify what you want to see in your cross correlation matrix.

As before you can have the matrix display either the matching quality of the current offsets or the quality of the best matchings. You can also sort the matrix rows and columns by a mean of best or current offset correlation values towards other members.

This is an excellent method to find a suitable member to start with when you want to synchronize an unsynchronized collection.

By having the matrix cells colored (chocolate graphics) according to the matching quality, you can easily find the best - and the worst - matching cases.

Using a .rwl or .fh file vs using a .fil file when saving a collection

Question: I have created my collections as .fil files. When and why should I have them stored in .rwl or .fh files?

The .fil file format allows me to collect a number of different coordinate (or ring width) files and relate them to each other with their offsets after crossdating. When finding errors I can easily change the inner content of a coordinate file without affecting its membership in the collection. I can backup my whole datastructure on a CD to keep in an archive. So why should I use .rwl or .fh files?

Answer: The answer is "communication" both with other projects of yours and communication with collegues and for publishing in the ITRDB.

When you open a .fil collection with 50 member files, the program looks up all these files. If you have changed the beginning of any file, that file will not be crossdated correctly to the rest of the collection. If files have been moved to another directory (that can certainly happen after two years) the collection will not open as expected.

If you move the collection to another directory, you need to have all your coordinate files placed around it as before to be able to open the collection.

On the other hand, if you Write the collection as a Tucson .rwl (or as a Heidelberg .fh) file that file will be a self-contained entity. All tree ring data is within it. You can send it anywhere, you can copy it, you can do what you like. Data will not get lost. And still you can manipulate the members. Though you do not have the coordinate information and images available for update. But that is usually not necessary during a later stage of a project or when a reference is being "just used".

As a bonus, opening a .rwl or .fh file is much faster than opening a .fil file, which matters for very big collections especially if you only want to create a mean value of the collection.

When saving a collection as .fil, also unchecked members can be included together with their checking status

When you save a collection in .rwl or .fh format, only your checked members will be written out.

When you save a Collection in .fil format, you can either save only your checked members or you can save the collection with all members with their checking status preserved. This gives you a way to save your job e.g. before you have finished it.

Statistical reports
Test towards rest of collection when multi radii samples are present.

The test towards rest of collection gives a list of how well each member of a collection matches towards the rest of the collection. When some members actually consist of several radii like NMG01A, NM01B then you will get misleading results if NMG01B is included in the rest when we calculate how well NMG01A matches to the rest.

When "Sum by stem" is checked, all radii of the same stem will be excluded from what is considered as "the rest" of the collection. "Sum by stem" is checked by default when a collection is opened.

Mean value and Standard deviation of Test towards reference and Test towards rest of collection.


NMGE4B      55     -5  0.70  6.7   49       -5  0.70  6.7   49 1995  ...
NMGE5A      54     -9  0.66  5.7   44       -9  0.66  5.7   44 1995  ...
NMGE5B      38      8  0.52  3.6   37        8  0.52  3.6   37 1995  ...
NMGE6A      40      2  0.52  3.7   39        2  0.52  3.7   39 1995  ...
Mean corr. of first column when overlap >= 40 (54 samples): 0.50
Standard deviation intervall 0.33 - 0.68

The report ends with statistics on how well each member matches towards the reference used.
In the example above a collection based on Norway spruce (PCAB) from Nämdö has been compared towards a reference solely based on Scots pine (PISY) also from Nämdö. The report shows a high probability that an individual spruce sample will match towards the Scots pine curve with a correlation coefficient in the range 0.33-0.68

When the same calculation is done for living trees of Spruce and Scots pine from Dalarna in Sweden (Axelson) the range is only 0.21-0.36. When based also on timber from old houses, the range is as low as 0.11-0.31

A few more old tips

Exporting ring width data from Excel

A tip from Wayne Hamilton:
Sometimes I create ring width files in Excel and then decide to analyse them in CDendro. Because CDendro expects to open a .wid file, it is necessary to convert the Excel file to Example.wid.  I do this by copying a column of ring width data from Excel to Notepad. [The top of the column should be the youngest ring.] Then I "Save As" the Notepad file as Example.wid and select "All Files" as the type, placing the new file into the appropriate folder.

A comment by Lars-Åke:
My version of Notepad has no "All Files" alternative, so then I have to save as "Text document" (i.e. the file is saved as Example.wid.txt) and then remove the .txt extension from the file name afterwards. (See also tip below on handling file name extensions in Windows.)

A tip on handling file name extensions when saving data from Internet Explorer:

Question: I've been going through the training sessions but find that I am having problems downloading the pola014.rwl file .. It automatically saves as a html through the 'save as' function and if I copy and paste into Notepad I can only do it as a txt file, which when changed to .rwl it saves as .rwl.txt and I cant get rid of the txt .... so the programme will not open the file ..

Answer:
One way is like this:  (best way!!!)
When you see the LIST of various files (at the ITRDB), right-click on the file you want to store.
In the pop-up menu choose Save Target As
In the Save-As pop-up menu change SAVE AS TYPE from Text document to ALL FILES
This will save the file with its original extension and not the added .txt extension.

If you are farther on and are seeing the actual data on your screen: (i.e. you have clicked the file name in the list mentioned above)
Select from the top-bar menu File/Save As
See that Save as type is set to Text file (.txt)
Save it by clicking the Save button.
You then have to change the file name from pola014_rwl.txt into pola014.rwl using your Windows Explorer.

Another solution:
When you see the data on you screen, select the whole text (e.g. press Ctrl-A) then do a copy to the clip-board (Ctrl-C)
Then go into the Notepad editor (Programs/Accessories) and paste the text into it (Ctrl-V).
Use the menu-command File/Save As and select Save-As-type to Text document
If there is an alternative "All files", then select that!!!! (no such alternative in my version of Notepad)
If there was no All-files-alternative, you have to afterwards remove the .txt extention.

You can do that already from Wordpad if you use the Save-As command again.
Then you can see the filename in the list as pola014.rwl.txt
Right-click on it, and select Rename and remove the .txt extension.
You will then probably get a warning - just ignore it and then exit the Save-As dialog box through the Cancel button.



Printing Workbench correlation reports and reports on collection properties

Question: I have not been able to find a good way to get the reports to print out so that the columns are lined up.  I can right click on the report and it will print correctly (which is very convenient) but it will only print the portion of the report that is visible and the part that is not visible is not printed.

Answer:

Reports in a collection's report frame:
Use the menu command: Collections/Save collection report As (.txt or .htm) to store the report in a separate file. This can be edited or printed from MS Word.

You can also use a Copy-text/Paste-text operation if you first paste into e.g. the Notepad program. If then copied from that program and pasted into e.g. Wordpad or MS Word using a Courier typeface, the result looks  fine.

Correlation reports created from the "Workbench" of a sample:
Select the text, copy it (Ctrl-C) and Paste (Ctrl-V) it into e.g. the MS Word editor. See that a font with fixed width is selected, e.g. Courier.

To select ALL text: Click to the left on the top line of the report.
Scroll down to the end of the report. Press the shift-key and hold it down and click after the end of the last line. The whole report text then lights up, i.e. it is selected! Press Ctrl-C to copy it!

On reading too small text

Many sites on the net are made with very small fonts - I often have problem to read the text. In most Browsers you can click the "Ctrl +" key to make the text instantly bigger without changing any other settings. And make it smaller with "Ctrl -".
Best use the control key in combination with the + and - keys to the very right of your keyboard.


Copyright © 2014, Cybis Elektronik & Data AB, www.cybis.se