ClustanGraphics8 (2005 release) offers the following features:**Read Data ** Reads a data matrix of any size. For example, a
million cases by 5 variables, and 500 cases by 40,000 variables, on any Pentium PC. Suitable for data mining and gene expression studies. Missing values are now allowed. See reading data. Can now be used with mixed data types.
**Compute Proximities
** Computes a proximity matrix by several proximity measures for up to about 5000 cases. Accepts missing values, mixed data types and computes binary and
continuous similarity coefficients, including Gower's General Similarity Coefficient. See computing proximities.
**Read Proximities
** Reads a proximity matrix from another package, spreadsheet or file. For example, you can
cluster variables on a correlation
matrix to produce a discrete factor analysis. Accepts missing proximities. See reading proximities.
**Clustering Proximities **
Hierarchical cluster analysis on a proximity matrix by
11 clustering methods. Detects and tests sensitivity of tied proximities. Allows for missing values, either in data or proximities. See clustering proximities. Can now be used with mixed data types.
**Cluster Exemplars ** For any hierarchical cluster analysis, finds the typical members, or exemplars, of each cluster. Used to identify and label clusters. Allows for missing values. See
cluster exemplars.
**Re-Order Tree **
Finds an optimum case order for any proximity matrix and hierarchical cluster analysis. Helps to compare clusters, and reveal their key defining characteristics. Allows for missing proximities. See re-order tree.
**Cluster Profiles **
Charts the key characteristics that differentiate between the clusters at a given level. Order by variables, or by clusters. Allows for missing values. See cluster profiles.
**Cluster Data
**
Unique, fast hierarchical cluster analysis directly on a data matrix. Has been used to cluster 120,000 cases hierarchically on a Pentium III notebook. Can now handle missing values, and differential case or variable weights. See cluster data.
**Cluster Keys ** Fast hierarchical divisive cluster analysis working directly on a data matrix. This procedure
finds the key variables to partition a dataset. Has been used to cluster 100,000 cases hierarchically on a Pentium III. See
cluster keys.
**Density
** Seeks clusters of cases in areas of high density and tries to find natural clusters. Hierarchical version of Mode Analysis.
**k-Means** Optimizes the
Euclidean Sum of Squares at a given cluster level by k-means analysis, or iterative
relocation. Use it to build a summary cluster model for very large datasets. Has been used to cluster a million cases. The search algorithm is optimized and guaranteed to converge. Can now be used with mixed
data types.
**FocalPoint Clustering** Two-stage k-means clustering method with six starting strategies and three
ending strategies. Outliers and intermediates can be removed from the cluster model, and the top solutions can be saved for profiling and evaluation. Clusters around exemplars or means.
**Multidimensional Scaling**
finds new continuous variables corresponding to the proximity matrix, allowing the relationships between cases and clusters to be displayed on a scatterplot, to aid graphical visualization of cluster analysis results. See ClustanMDS.
**
Principal Component Analysis** reduces the dimensionality of the input data, such that the first two
principal components correspond to the best fit of a graph through the data. Most pca programs only handle continuous variables, but our pca will also cope with mixed data types and
missing values and is very efficient, capable of handling hundreds of variables. The resulting relationships between cases and clusters to be displayed on a scatterplot
, to aid graphical visualization of cluster analysis results. See ClustanPCA.
**Outlier Analysis ** Identifies outlying cases by reference to a classification and removes them to an unclassified set. Used to calibrate a cluster model. Allows for missing values.
**Classify Cases **
Finds the nearest cluster for an unlimited number of new cases by reference to a specified cluster model. Handles missing values. See classify cases. **Truncate Tree ** Reduces a large tree to manageable proportions, computing cluster means and
proximities between clusters. Allows for missing values.
**Display Scatterplots
** Plots cluster scatterplots for any input variables, or for illustrative external variables such as principal components. See cluster scatterplots.
**Navigate Tree **
Top-down summary tree for a cluster model which can be displayed, edited for publication, pruned, or used to display cluster profiles. Use with k-means to classify and summarize large datasets. Can
now handle missing values. See navigate tree.
**Best Cut **
Seeks the optimum partition of a tree by significance tests.
**Tree Styles **
Different ways of displaying a tree. Options include colour shading, case labelling, spacing, physical dimensions, line, orientation, font and zoom. See display tree.
**Shading Proximities **
Displays a shaded representation of any proximity matrix, using colour to differentiate between-cluster and within-cluster proximities. Allows for missing proximities. See displaying proximities.
**Weighting **
Variables and cases can be given differential weights. Variables can also be transformed to z-scores or unit ranges.
**Labelling** Cases, variables and clusters can be given alphabetical labels of any length.
ClustanGraphics offers a very interactive "point-and-click" Windows user interface. Help is provided by a standard Help file, floating hints and context-sensitive help. The Clustan Wizard simplifies a hierarchical cluster
analysis with standard defaults in a single dialogue... *three clicks and you're clustered*. ClustanGraphics is very easy to use, requiring no prior professional knowledge or training. If you haven't already done so,
Preview or check out Clustering Large Datasets
. Oh, and don't forget! ClustanGraphics includes a 64-page illustrated Primer.
Unfortunately we can only give you a superficial tour of what you can do with ClustanGraphics in this webpage. To order ClustanGraphics on-line click **ORDER** now. |