|
ConClus - Constrained Cluster Analysis |
Last update: 26. April 2006
ConClus is a program that can perform cluster, factor, and scale analyses,
some of these cannot performed with other statistical programmes.
ConClus is made for huge data sets, many
variables and many cases. You can weight cases and variables simultaneously,
and you can perform a confirmatory cluster analysis. This method enables you to
test a cluster configuration that you expect, and ConClus will test whether this
configuration is in your data or not.
Features of ConClus
- Cluster analysis with up to 2000 variables, 45 clusters and 1 million cases.
- Confirmatorical cluster analysis: testing of a cluster structure
using a second data set or testing of theoretical hypothesis.
- Canonical cluster analysis: The whole data set is split into 2 splits.
The clusters are extracted for both splits seperatly, but simultaneously.
Both solutions are calculated for a maximum correspondence.
The cluster structures are tested inference statistically
whether they are identical or whether they can be reproduced inadequatly.
- Processing of missing values: iterative substitution of cluster averages
instead of casewise deletion.
new: test whether missing values occur
significantly in certain clusters.
- Inclusion of restrictions: a wide number of theoretical restrictions can
be modelled, e.g. the range of some - or all - variables can be
defined within single clusters.
- Weighting of variables: polytomious variables, e.g. the religion of
an interviewed person, can be transformed into weighted dummies.
Relevant variables can be weighted stronger.
- new: factor analyses: principal component analysis with ULS and orthogonal
varimax rotation for up to 500 variables. The regression method is
used for the calculation of the factor values.
- new: scale analysis with 100 variables and ULS iteration of the communalities
(Jöreskog), Cronbach's α, the explained variance and the highest
possible reliability are calclated.
- new: weighting of cases, e.g. to process aggregated data.
- Very fast due to the integration of elementary AI-algorithms: ConClus saves
a lot of computation through learning (by recognising the data
structure and elimation of useless solutions of equation systems).
- Simple operation and control language.
- New manual (in German) with over 160 pages and a lot of examples showing applications and interpretation of the results.
- real 32-bit software, very fast.
- ConClus runs under Windows XP in MS-DOS mode