Y o u could click New Data and read your data from a space delimited, character sequence, binary or Excel file; you couldtransform the variables, or specify weights and labels. But to get started quickly, we suggest you open a ClustanGraphics file which was created earlier, for example Mammals.cls. Then click OK and a tree for our Mammals Milk case study will be displayed.
Now point and click the tree. ClustanGraphics shades the cluster partition you selected according to a vertical section of the tree. Congratulations! You just specified your first cluster model - in this example, 5 clusters of mammals based on the composition of their milk. Note that each cluster has anexemplar, or most typical case, which is underlined.
Click Copy and your tree will be copied into a report, presentation, or website. That's exactly what we did here. Pretty easy, wasn't it! With about six clicks we loaded our Mammals case study, specified a 5-cluster model and published it here.
There's an important detail we should mention here. A tree obtained by hierarchical cluster analysis can be displayed in many different ways, because the order of the cases within any cluster can be reversed. Our tree has beenoptimally re-ordered to maximize the rank correlation of the proximity matrix. So the order of the cases and clusters in our Mammals tree is meaningful, more so than the trees produced by other clustering programs.
ClustanGraphics also allows you tonavigate and edit your tree, using a unique vertical representation that's ideal for presenting summary model results.
You can view a cluster model on a scatter diagram.Cluster Scatterplots shows an example of the 5-cluster model for Mammals, illustrated above.
Now point at any cluster and click the right mouse button. A floating menu pops up beside the selected cluster with options for exploring it further. In the example we clicked cluster 4 (Dog-Rabbit). The popup menu shows that it has 6 members and that Deer is thecluster's exemplar. We can now obtain a list or table of the cluster's members, explore its profile or display cluster scatterplots.
ClickingProfiles reveals our unique ClustanGraphics Profiles chart.
The example displayed below lists the mean percentages of Protein in the milk for each cluster. It shows that the selected cluster 4 (Deer) has the highest average protein, with cluster 5 (Seal) a close second. Click the chart and it advances to the next variable. Click Pivot, and the chart changes to show a summary for each cluster in turn.
Now click Table to view the full table of cluster means or copy it into a report, presentation or spreadsheet. Perhaps you'd like to try clustering by another method - ClustanGraphics has 11 methods of hierarchical cluster analysis. ClickCompute Proximities to see how you can compute a new proximity matrix on your data and cluster it for up to 10,000 cases.
If you're involved in analyzing large surveys, rest assured that ClustanGraphics can cluster many thousands of cases.Clustering Large Datasets shows how we can construct a tree for 120,000 cases by Ward's Method, using a basic Pentium III PC. It's a very efficient method, unique to ClustanGraphics.
Another approach to clustering large datasets is usingk-Means Analysis. Our k-Means preview shows how you can refine a large tree by reassigning cases and deleting outliers. It can then be truncated to the model that interests you, and used to classify new cases. Alternatively you can specify the initial clusters, or assign them randomly, and then iterate to an optimal cluster model. We have run k-Means Analysis on a million cases, when our exact relocation test for the Euclidean Sum of Squares becomes very important.
No other software can match this for performance! To find out more about ClustanGraphics please click the toolbar buttons in the left margin. Or seeCase Studies for further examples.
lists it's full range of functionality. If you are a Clustan/PC user, then Clustan Comparison answers the question: What's the difference between Clustan and ClustanGraphics?
ClustanGraphics is fully described in our 60-page To order ClustanGraphics on-line click
ClustanGraphics is fully described in our 60-pageClustanGraphics Primer, an introduction to cluster analysis for new users. No prior statistical knowledge is assumed. So the ClustanGraphics Primer is ideal for student classes, and very affordable if you have a network or site license. One copy is supplied free with each ClustanGraphics purchased.
To order ClustanGraphics on-line clickORDER now.