Nicht angemeldet

Home | Datenbank | Biologie | Aquaristik | Software | Registrierung

CLORD, a program for Classification and Ordination

The special feature of this statistical software is the graphical, interactive design. In various windows exists the possibility of zooming areas of dendrogram or scattergram with the mouse and saving partition-tables of agglomerated data. The order of the objects and attributes is always based on the sequential carried out multivariate analyses.

It doesn't matter if you are interested in the analysis of rows or columns or both. A single mouse click changes the interpretation of the data, and it is possible to perform different transformation on each action.

The original data is hold with each window; the titles inform about the filename, transformations and methods. Multiple datafiles can be load at one session, the data selection and order is always taken from the previous active window.

The multiple document interface lets you easily compare different data or the same data with different methods. There is the possibility of minimize some windows as icons and tile or cascade the others.

During longer calculations other applications can run, you are informed about the progress with the main window title. High resolution printing is available with all Windows printer drivers, further processing of the graphics in other programs through clipboard support.

Glossary

Agglomeration

When two objects are merged in the hierachical clustering method, their similarity to the other objects have to be calculated new. The kind of this calculation determines the agglomeration method.

Attributes

Attributes usual describe the properties of the objects to be analysed. They are measured on an ordinal, interval, rational, or qualitative (for example colours) scale. It is also possible to analyse the attributes (variables) in a so called R-analysis (only similar to a Row-analysis when they are listed as rows in the original datafile).

Block clustering

This analyse both the attributes and the objects so that in the ordered data matrix similar values appear as blocks.

Classification

With a classification the objects are divided in discrete groups.

Data matrix

The data is described by row and column labels and the corresponding values. Thus, it doesn't matter whether the objects stand in the rows or in the columns.

Dendrogram

A dendrogram is the graphical representation of the structure obtained in a hierarchical cluster analysis, also called tree. At the left side the similarity is the largest and each object itself forms a cluster whereas at the right side all objects are merged in one big cluster.

Dissimilarity coefficient

With this class of resemblance the most similar objects have the largest, the most dissimilar objects the smallest values (for example euclidean distance).

Hierarchical clustering

In a hierarchical clustering it is possible to obtain an optional number of clusters. The range  goes from the whole number of objects to only one big cluster and is determined by a cut value for the similarity of the one after the other merged clusters.

Misssing values

In most applications attributes with missing values (not possible to determine) are completely let out of the analysis, but sometimes substituted by approximated values. In this program version you must supply the handling with previous treatment of the datafile.

Objects

Objects are in most applications the items to be analysed (classified or ordinated), described through attributes (variables). In a statistical sense this is also called Q-analysis.

Ordination

With a ordination objects are ordered along one or more hypotetical continuous gradients.

Resemblance

The resemblance matrix can be calculated of either a similarity or a dissimilarity coefficient. It has a diagonal form because the resemblance between, for example, object 1 and 2 is the same as between 2 and 1.

Scattergram

In a scattergram the object are plotted in a twodimensional coordinate system with their first two component scores.

Similarity coefficent

With this class of resemblace the most similar objects have the largest, the most dissimlar objects the smallest values (for example correlation).

Transformation

A Transformation changes the values in the data matrix either based on vectors (columns, rows) or by each entry alone (Logarithmisize, Power Transformation). Thus different scaled attributes can be compared.

Anderberg, M. R. (1973) Cluster Analysis for Applications. Academic Press, New York.

Dale, M. B. (1989) Similarity measures for structured data: a general framework and some applications to vegetation data. Vegetatio 81: 41 - 60.

Lance, G.N. and Williams, W.T. (1966) A general theory of classificatory sorting strategies. I. Hierarchical systems. Comp. J. 9: 373 - 380.

Lance, G.N. and Williams, W.T. (1967) A general theory of classificatory sorting strategies. II. Clustering systems. Comp. J. 10: 271 - 276.

Noy-Meir, I. (1973) Data transformations in ecological ordination. I. Some advantages of non-centering. J. Ecol. 63: 329 - 341.

Pirktl, L. (1983) Probleme und Algorithmen der Clusteranalyse. Dissertation, ETH Zürich.

Pritchard, N. M and Anderson, A. J. B. (1971) Observation on the use of cluster analysis in botany with an ecological example. J. Ecol. 59: 727 - 747.

Romesburg, H.C. (1984) Cluster analysis for researchers. Lifetime Learning Publications, Belmont.

Sneath, P.H.A. and Sokal, R.R. (1973) Numerical taxonomy. W. H. Freeman, San Francisco.

Steinhausen, D. und Langer, K. (1977) Clusteranalyse. Walter de Gruyter, Berlin, New York.

überla, K. (1971) Faktorenanalyse. Springer, Berlin.

The following copy of CLORD is distributed without any warranty. Please extract the downloaded Zip file into a directory of your choice and start 'clord.exe' by doubleclick in Windows- or NT-Explorer.

Download now

Home | Gästebuch | Zum Seitenanfang