Dendrogram
Look under the "Help" tab above for the Quick-start guide.
Load different data
Load a different data set by clicking one of the links in the table below.
The current data
Below are details of the currently loaded data set:
The largest merge height (dissimilarity value) is
0.
There are
0 leaf nodes.
Groups data will be displayed here.
Here will appear output in response to formation
of groups using the green dragable threshold bar and the "Apply threshold"
button, or the "Set no.of groups" toolbar.
Dendrogrammer v 1.0.1
The Dendrogrammer was created as one component in the analysis of
multivariate data. In particular it is intended to allow already
pre-processed data to be
- classified into groups and then either
- used in further analysis, or
- used in some form of presentation about the data.
Current Features
- Dendrogram construction
- File read
- File read now includes a manifest file which cues a clustering data file
(node data) and a leaf description data file (leaf data).
- File read now is integrated in the application page. Hard coded data
loads if no data files are found.
- Window width optimisation
- Group formation by adjustable threshold
- Cluster click/tap reports into dialog
- Group formation reports with dialog, reports into separate tab,
that tab pulses to notify. The pulsing stops when the tab is clicked
- The threshold is a dragable element on the graph
- The threshold can be adjusted also using toolbar buttons
- The desired number of groups can be entered and it will set
the threshold to fit
- Leaf labels on the vertical axis. These are interactively styled in response
to events.
- Dialogs now guaranteed to be under 75% of window ht, scrolling if needed
and using a default or wider width as appropriate on PC platform.
Scrolling dialogs are not deployed on tablet platform as these are
not suitable for tablets.
- The group banding (and groups) can be removed with the "Remove" button
- Summary dendrogram: For use in dendrograms with a very large number of
leaves. Effectively truncating the dendrogram up from 0 ht to a level at
which the structure can be discerned and then the clusters interrogated,
rather than having hundreds of leaves along the bottom.
- The Descendant Nodes Dialog contains a search link. The URL for this is
specified in the .dat file which loads the data. The search link passes
the cluster's leaf terms to a search engine.
The technologies deployed in the application include
- JavaScript - The major part of the application is written in JavaScript
- Raphael Javscript library
(by Dmitry Baranovskiy)
- Used to make the dendrogram graphic elements. Raphael is truly amazing.
- (Indirectly) SVG and VML - Raphael uses SVG or VML to do the drawing.
Raphael switches seemlessly depending on the browser
- jQuery - For some of the cross browser Dynamic HTML
- jQuery UI - For some of the user
interface supporting the application
- HTML5 - Although there is nothing especially HTML5 about it really.
- PHP - the data files are loaded using PHP.
The Dendrogrammer was created
- by David Allan Robb
- for his MSc project at Heriot-Watt University
- in summer 2011
Quick-start Guide
- In a nutshell-
- Drag the green bar (Threshold bar)
to adjust the dissimilarity threshold for group formation.
Then click the "Apply threshold" button to form groups.
- "Remove" will get rid of the groups.
- Or enter the desired no. of groups in the "no. of groups field" and
click the "Apply no. of groups" button.
- To create a summary dendrogram, drag the threshold, then click
"Truncate".
- "Restore" redraws the original dendrogram.
- Another way to adjust the threshold is to enter a number in the
"Threshold value field".
Help
- Dendrogram tab - View and interact with the dendrogram made from the clustering data.
- Click or tap a cluster (element on the graph).
Then read about that cluster in the dialog that appears.
- That dialog also contains a search link. The URL for this is specified
in the .dat file which loads the data. The search link passes the
cluster's leaf terms to a search engine. It should open in a
browser pop-up window or browser tab.
- Adjust the green threshold bar and click the button to form groups of
leaves (data points)
- See more detail about the groups you form under the "Groups" tab.
- The threshold can be adjusted by click-and-drag on the green bar,
or using the threshold toolbar buttons, or entering a number in the
"Threshold value field".
- If you are using a tablet computer then drag on the green bar
with your finger. If you find you can't drag the bar accurately then
use the toolbar buttons to move the threshold. You might find that
after moving the threshold bar away from the max with a button then you can get
a better feel for how it drags using your finger.
- If using the toolbar fine adjust buttons (plus and minus), note that
the amount of movement with minus is set slightly less that with plus.
This allows even more fine adjustment if you alternate from
plus to minus around the point you are looking for.
- To specify a given number of groups, enter the desired number in the
field next to the "Apply no.of groups" button and click the button.
The threshold will be chosen automatically for you.
- To view a summary dendrogram (or truncated dendrogram),
drag the threshold to some point below which you wish to discard
the leaf detail. Then click the "Truncate" button. The resulting
summary dendrogram will have everything below the threshold
removed and be reshaped to fit the previous space. The origin is at the
dissimilarity ht. you chose with the threshold bar. Tapping/clicking
a cluster will still report all the leaf detail as before,
despite the leaves being invisible. Groups can be formed in the same
manner as before.
- The "Restore" button will restore the original full dendrogram and
the leaves as it was when you first viewed the current data.
Loading new data or the same data again will have the same effect,
as will reloading the page with your browser.
- Use Control and + or - to zoom in and out ( or pinch on a tablet ).
Look in your browser's View menu for more zoom controls.
- What equipment to use: The dendrogrammer was designed to be cross platform
(works on PCs with mouse and tablets with tap)
- What browser to use: The dendrogrammer was designed to be cross browser. It
is written in JavaScript (with a little bit of PHP to load the data). So it
works in pretty much any browser that has Javascript enabled. However it is
my experience that if one considers the latest browsers, the JavaScript
interpreters in Firefox (Gekko with Tracemonkey), Safari (Webkit with
Squirrelfish), and Chrome (Webkit with V8) are significantly faster (e.g.
by a factor of 2 and more) than the
"JScript" engine in Internet Explorer 8 and below (However, IE9 32-bit is
good, but IE9 64-bit is reputed to be slow).
So if one is seeking faster smoother performance with this particular
application then one might wish to use Firefox, Chrome, or Safari
(as at July 2011).
- Data tab - Load different data or find out about the current
data used to make the dendrogram.
- A table of links is shown representing all the available data sets.
Clicking a link loads that data set and makes a dendrogram from it.
- The currently loaded data set is described and displayed in 3 tables
- One table shows the contents of the .dat (or manifest) file. This
contains meta data about the data set: the names of the two data files
and a description of the data set.
- The next table shows the contents of the clustering data file (or node
file). This should be a csv file generated by MATLAB. It should have
3 columns. The first two columns are cluster or leaf designation
numbers. The third column contains the merge height or similarity
measure for that cluster.
- The last table shows the contents of the leaf description data file (or leaf
file). This is a csv file. The first row describes leaf number one. The second
row describes leaf number two and so on. It should contain two columns. Column
one should hold the short leaf descriptions or labels. Column 2 should contain
a longer description for each leaf (this can be a repeat of the short label).
- Loading a new data set or the same data set again will result in a page
refresh and will clear any displayed groups or summary dendrogram which you may
be displayed on the Dendrogram tab.
- Adding further data sets can only be done by someone with write-access
to the web space where the application resides (See 'Documentation'
in the About tab.)
- Groups tab- View the detail of the groups formed using the threshold adjuster
and toolbar. The detail can be copied and pasted into other software for
further analysis or presentation.
- About tab - Read about the application.
- Help tab - This page you are reading now is the Help.
Zoom Help
If you find the leaf labels too small you may wish to zoom in.
If you find your screen height too small to fit a large dendrogram
you may wish to zoom out.
- How to zoom on a PC or laptop
- On most browsers Control and
+ will zoom in.
- Control and -
will zoom out.
- Control and mouse
wheel works too on most browsers.
- Look in your browser's View menu for more zoom commands. E.g.
Some browsers offer a separate zoom for text only. If using that,
bear in mind that your browser sees the entire dendrogram including
leaf labels as graphics.
- Once you have zoomed, refreshing the page will adjust the dendrogram
to fit the width you have chosen.
- On a tablet device
- Use the open or close pinch gesture to zoom.
- Once you have zoomed in, as you would expect, refreshing the page
does not affect the dendrogram size. It remains zoomed until you
zoom back.
- Summary Dendrogram feature
- By setting the threshold bar and clicking "Truncate" you can
reduce the amount of leaf detail displayed and reduce the height of
the dendrogram.
- Clicking a cluster on a summary dendrogram still reveals all of
its descendant leaf detail in the dialogs. Group formation works
as normal.
- The "Restore" button restores the normal dendrogram.