Multilevel model

Properties of social networks
Mean network size was 80.67 (SD = 39.84,
Mdn = 75), with a range of 10–279
(Figure 1). The shape of the distribution is
similar to that found in other studies of social
networks (Bernard et al., 1990; Hill & Dunbar,
2003). The size of the kin and friend
networks is given in Table 1.
Multilevel models
The results of the two models are presented
Table 2. The intraclass correlations (ICCx)
of the models were 0.18 for Model 1 and
0.24 for Model 2. An ICC < 0.50 indicates
that there was more variation in time to last
contact at the level of the network member
Network size
0 20 40 60 80 100 120 140 160 180 200 220 240 260 280 300
Number of cases
Figure 1. Distribution of personal network
than at the level of the participant. However,
the values of ICC demonstrate the need for
multilevel analysis as there is clearly variation
between participants on time to last contact,
and thus simply analyzing the data without
taking into account that network members are
clustered within participants would result in
an increased Type 1 error rate (Tabachnick &
Fidell, 2007).
Face contact (Model 1)
Level 1 effects. Participants had a shorter
time to last contact to network members at
high levels of emotional closeness and to network
members living closer by. Participants
had a shorter time to last contact for friends
as compared to kin. Participants did not have
a shorter time to last contact to female, as
compared to male, network members.
Level 2 effects. As predicted (H1a), kin network
size had a positive effect on time to last
contact. Thus, participants with larger kin networks
had a longer time to last contact with all
network members (including both friends and
kin), as compared to participants with smaller
kin networks. However, contrary to H1b, there
was no significant effect of the friend network
size on time to last contact. H2 was supported:
The mean...
tracking img