Social Network Types
Eight criterion variables were used in
K
-means cluster
analysis, namely, total network size, frequency of contact
and support exchange with immediate kin, frequency of
contact and support exchange with distant kin, frequency of
contact and support exchange with non-kin, and engage-
ment in social activity. Standardized scores were used to
eliminate scaling difference among variables. Because pre-
vious studies have shown consistently a range of four to six
clusters in older adults, we began by specifying four clus-
ters and then gradually increased the number of clusters.
The choice of appropriate cluster formation was determined
on the basis of three criteria: (a) eta square from multivari-
ate analysis of variance (MANOVA) showing the strength
of the relationship between clusters and the set of criterion
variables, (b) number of cases in each cluster, and (c) mean-
ingfulness of the formed clusters. On the basis of these cri-
teria, a fi ve-cluster solution was found to be most suitable.
Consistent with previous studies (e.g., Fiori et al., 2006 ,
2007 ), we found four major network types: diverse, friend
focused, family focused, and restricted. On top of these,
we found a fi fth network type (labeled as distant family)
Social Network TypesEight criterion variables were used inK-means clusteranalysis, namely, total network size, frequency of contactand support exchange with immediate kin, frequency ofcontact and support exchange with distant kin, frequency ofcontact and support exchange with non-kin, and engage-ment in social activity. Standardized scores were used toeliminate scaling difference among variables. Because pre-vious studies have shown consistently a range of four to sixclusters in older adults, we began by specifying four clus-ters and then gradually increased the number of clusters.The choice of appropriate cluster formation was determinedon the basis of three criteria: (a) eta square from multivari-ate analysis of variance (MANOVA) showing the strengthof the relationship between clusters and the set of criterionvariables, (b) number of cases in each cluster, and (c) mean-ingfulness of the formed clusters. On the basis of these cri-teria, a fi ve-cluster solution was found to be most suitable.Consistent with previous studies (e.g., Fiori et al., 2006 ,2007 ), we found four major network types: diverse, friendfocused, family focused, and restricted. On top of these,we found a fi fth network type (labeled as distant family)
การแปล กรุณารอสักครู่..
