Fig 1.
The process of link-tracing data collection, by month and location.
The lines are indicating the time variation of the conducted interviews (red line for the Dâmbovița site, while blue for the Castellón site).
Fig 2.
Multi-layered network (the network of networks) built by the link-tracing sampling method.
(A) Inter-connected personal networks, through link-tracing ties and shared alters. (B) Zoom in on the personal network of respondents (the dotted area of A in the panel). Node colors designate places (countries) wherein individuals currently live, i.e., ‘blue’ for Dâmbovița (Romania), ‘red” for Castellón (Spain), ‘yellow’ for other places than Dâmbovița and Castellón (these other places could be other regions in Romania and Spain, or even other countries–for simplicity and illustrative purposes, in this example, we decided, by ‘yellow’, to mark other countries. Node shapes designate classes of individuals (alters) elicited by the fixed-number alter name-generators: ‘squares’ designate ego’s family members, ‘triangles’ designate ego’s friends, while ‘circles’ designate ego’s acquaintances. The ego (the respondent) is marked by a trapezoid shaped—black bordered node. Ego’s alters (family members, friends and acquaintances) who lived in Castellón (Spain) and returned to live in Romania are marked by ‘red’ bordered shapes (e.g., returned family members are marked by blue squared nodes with red border). A sub-set of nine alters was randomly sampled from the elicited set of alters, and the alter-alter existing ties represented. This is illustrated by the green edges. If two nodes are connected by an edge, that means the two nodes know each other and can contact each other independently from the ego. As all the nodes represent ego’s alters, the respondent (the trapezoid shaped node) is connected to everybody through a tie (see the thin gray edges). Orange edges indicate the cases wherein two egos share alters. The black arrows indicate the direction of the link-tracing sampling: an arrow’s origin marks the referee while the head of the arrow marks the referral. The plots were built using visone [57].
Table 1.
Methodological differences between NSIT and ORBITS studies.
Table 2.
The demographic profile of the initial subjects in the sample (the seeds).
Table 3.
Distribution of participation by referee’s gender and place of living.
Table 4.
Major demographic characteristics of the link-tracing network participants (survey respondents).
Fig 3.
Hairball visualization of the link-tracing network.
Node colors indicate three classes of nodes: red down-triangles–seeds (individuals living in Castellón, Spain), red–people interviewed in Castellón, Spain, blue–people interviewed in Dâmbovița, Romania. The network is directed indicating the chain-referral structure of the relational data. Directed dyads are indicative of referees (the origin of the arrow) and referrals (indicated by the arrow head). Size nodes are proportional to their out-degree (the number of referrals provided). By dotted perimeters are indicated components (disconnected parts of the network). The plot was built using the UCINET 6.0.
Fig 4.
Pattern of referral nominations based on sex and residence.
The visualization illustrates the pattern of referral nominations, based on sex and place of residence. For instance, the female respondents living in Romania (Dâmbovița) (see the “from” axis) indicated as referrals (see the “to” axis): 135 and 57 males living in Romania (Dâmbovița), 27 females and 15 males living in Spain (Castellón).
Fig 5.
Boxplots illustrating individual distributions of EI-index scores split by sex and residence.
For every study participant who provided referrals, an E-I index score was computed, either taking into account sex or place of residence. E-I index is computed by a simple formula: for a specific participant the difference between the number of ties between groups and within group is divided up to the sum of ties (e.g., for a female participant, the difference between her ties sent to males and her ties sent to other females is divided up to the total number of ties).
Table 5.
Sex / place of residence E-I index scores computed in the link-tracing network.
Table 6.
Micro-level determinants of nomination tie patterns, in the link-tracing network of study participants.
Table 7.
The demographic composition of the link-tracing network chains.
Fig 6.
The network has 4,855 nodes, 5,477 arcs and 2,540 undirected ties. Colors indicate network nodes’ place of living, i.e. red (Castellón, Spain), blue (Dâmbovița, Romania), and green (other countries). The network data was visually encoded with visone (stress minimization node layout) [57].
Fig 7.
The network of networks (ties split on residence).
The hiveplot illustrates how the 4,855 nodes and the 5,477 nomination ties within the network of networks are partitioned on residence (countries wherein people live: Spain, Romania and other countries). Each node is positioned on the axis based on its rank (out-degree or the number of people they nominated), under the principle of first served (for equal degrees, nodes are placed based on the assigned numbering in the dataset). Magenta ties indicate within-country social connections (i.e. 1,524 ties connect people living in Spain, and 2,237, people living in Romania), whereas gray ties, between-country connections (i.e. 1,133 ties connect Spain and Romania, 223 ties connect Romanians living in Spain to Romanians living in other countries, and 360 ties connect Romanians living in Romania to Romanians living in other countries). The axis illustrating the “other countries” class of nodes lacks within-ties by design. Nodes assigned to this axis were not interviewed, but only nominated by participants (contacts they have in their personal networks, either relatives, friends or acquaintances). The axes are unstandardized to emphasize differences in the volume of nodes.
Table 8.
Number of alters nominated within the name generators, split by type of participants.
Table 9.
Structural characteristics for chain networks, link-tracing network and the network of networks.