Effects of group size and lack of sphericity on the recovery of clusters in K-means cluster analysis

Auteur(s)
Craen, S. de; Commandeur, J.J.F.; Frank, L.E.; Heiser, W.J.
Jaar
K-means cluster analysis is known for its tendency to produce spherical and equally sized clusters. To assess the magnitude of these effects, a simulation study was conducted, in which populations were created with varying departures from sphericity and group sizes. An analysis of the recovery of clusters in the samples taken from these populations showed a significant effect of lack of sphericity and group size. This effect was, however, not as large as expected, with still a recovery index of more than 0.5 in the 'worst case scenario'. An interaction effect between the two data aspects was also found. The decreasing trend in the recovery of clusters for increasing departures from sphericity is different for equal and unequal group sizes.

Publicatie aanvragen

5 + 0 =
Los deze eenvoudige rekenoefening op en voer het resultaat in. Bijvoorbeeld: voor 1+3, voer 4 in.
Pagina's
127-145
Verschenen in
Multivariate Behavioral Research
41 (2)

SWOV-publicatie

Dit is een publicatie van SWOV, of waar SWOV een bijdrage aan heeft geleverd.