User Tools

Site Tools


en:hier-agglom

Differences

This shows you the differences between two versions of the page.

Link to this comparison view

Both sides previous revision Previous revision
en:hier-agglom [2019/04/14 17:23]
David Zelený [Flexible beta linkage]
en:hier-agglom [2019/04/14 17:50] (current)
David Zelený [Flexible beta linkage]
Line 44: Line 44:
 === Flexible beta linkage === === Flexible beta linkage ===
  
-Flexible clustering algorithm, which allows influencing the level of chaining by modifying the parameter β (β ≥ -1 and β  < 1). With β = -1 the chaining is the lowest, with beta approaching +1 the chaining is highest (similar to simple linkage method). Optimal representation of distances among samples is for β = -0.25 (results similar to Ward’s algorithm). The method can be used with both metric and non-metric coefficients of similarity/​dissimilarity. +Flexible clustering algorithm, which allows influencing the level of chaining by modifying the parameter β (β ≥ -1 and β  < 1). With β = -1 the chaining is the lowest, with beta approaching +1 the chaining is highest (similar to simple linkage method, <imgref flex-beta>​). Optimal representation of distances among samples is for β = -0.25 (results similar to Ward’s algorithm). The method can be used with both metric and non-metric coefficients of similarity/​dissimilarity.
  
 +<​imgcaption flex-beta|Flexible beta linkage algorithm applied on Bray-Curtis distance calculated from Dune dataset; three alternative beta parameters result in quite different outcomes: beta = -0.95 (left), beta = -0.25 (middle) and beta = +0.95 (right).>​{{:​obrazky:​classification-flexible-beta.png?​direct|}}</​imgcaption>​
 ==== Effect of transformation applied on species composition data ==== ==== Effect of transformation applied on species composition data ====
  
en/hier-agglom.txt · Last modified: 2019/04/14 17:50 by David Zelený