cluster analysis

(redirected from Cluster analyses)
Also found in: Financial, Encyclopedia.

clus·ter a·nal·y·sis

a set of statistical methods used to group variables or observations into strongly interrelated subgroups.

cluster analysis

Etymology: AS, clyster, growing together; Gk, analyein, to loosen
(in statistics) a complex technique of data analysis of numeric scale scores that produces clusters of variables related to one another. The technique is performed with computer software or statistics programs.

clus·ter anal·y·sis

(klŭstĕr ă-nali-sis)
Set of statistical methods used to group variablesor observations into strongly interrelated subgroups.

cluster analysis

a statistical method of arranging a set of observations into sub-sets, each of which groups together those observations having similarities.


1. in epidemiological terms a naturally occurring group of similar units, e.g. animals which resemble each other, with respect to one or more variables, more than animals in different groups do, or a group of cases of a single disease in time or space.
2. assembly of claw and teat cups, as part of a milking machine.

cluster analysis
1. statistical methods used to group variables or observations into strongly interrelated subgroups.
2. a statistical analysis of the relationships between clusters in time and/or space.
cluster fly
see polleniarudis.
cluster sampling
see cluster sampling.
References in periodicals archive ?
To see if the clusters from the above-mentioned cluster analyses varied based on individual difference variables, the following analysis was conducted.
The same procedure (26 agglomerative hierarchical clustering procedures using the centroid method and Squared Euclidean distance) was followed with the cluster analyses per couple for the restaurant task (See Table 3).
The Mann-Whitney U-test with Bonferroni correction was used to determine statistical significance between paired comparisons of clusters segregated by the threshold obtained from the results of cluster analyses.
From the results of cluster analyses, initial MC was found to be the best sorting parameter.
The initial cluster analyses were run using Ward's method with squared Euclidean distance as the measure of similarity.
To allow more discrimination between profiles, hybrid two-stage cluster analyses were conducted across lifestyle and personality characteristics.
Obviously, no manager would knowingly analyse random data, but if data sets are not rigorously tested, the solutions from such cluster analyses (and other statistical techniques) can essentially be seen as devoid of meaningful structure.
Cluster analyses were also performed on confirmed and probable case data.