Please enable JavaScript.

Coggle requires JavaScript to display documents.

Cluster analysis (distance measures (Euclidian distance (image), Block…

- - - - Sum of the absolute distance between two variables
- - - - average linkage between all cases in new cluster as small as possible
    - - average distances of all pairs across clusters
- - - - Start with proximity matrix
      - combine two closest variables into one cluster
      - continue to combine variables/clusters at each successive stage
      - one large cluster is formed
      - Agglomerative schedule
        
        gives agglometration coefficents
        
        =distances at which clusters were combined
        
        expressed graphically through a dendogram
  - - - new cluster means generated with each new cluster formed
  - - - cases grouped into sub set of smaller clusters
        
        via method similar to k means
    - - sub clusters combined using standard hierarchical technique
    - - combines both methods
      - handles outliers
      - allows for both categorical and continuous measures
      - reseracher can either set no. of clusters or allow program to determine
    - - can depend on the order of the data input particularly with small samples