4. Diseño etnográfico.
4.2. Descripción del contexto e interpretación de los resultados.
i
n
fect
ious organism enters a social n ehvork, closeness describes for each i ndi v iduaL the"time-until-arnval " of the mfecti o us o rganism. assuming
that
i n fect i on takes the shortest path (Borgatti. 1 995). This means the more remote an md ividual possum is in the netvvorkthe
l ess l i kelv it is that it w i l l become infected. and also l ess able to i n fect others . Onthe
other hand a · central ' possum will be more likely to b ecome infected and will be verv e1Tecti ve at spreading i nfection through the network.The closeness centralisation index is a measure for the whol e group and it quantifies closeness over all i ndividuals
in
the group. Itis
the sum o f the differences in c loseness score between the most central member (highest indiYi dual closeness score) and each o fthe members
o f the group d ivided by the maximum c loseness score possib
le for the group. As the i ndex increases. i ndividual possums i n thenetvmrk
behm e less alike. that is,t
h ey are more heterogeneous . Itr
eaches its ma'i.imum \\ hen there is
one central mdi vid ual(maximum heterogeneity). and mini mum \Vhen al l i ndividuals in the group are equally
close
(Wasserman and
Faust, 1 998). In the context of disease transmission, this m eans thatin
anetwork
\Vith a high closeness central is ation i ndex. infecting the central i ndi-vidual will result i n efficient disease spread, \vhereas i nfectiOn in less central animals \Vi ll be lessI f social i nteract i ons are i nterpreted as representing
a
com
mu nicat
i o ns nenvork, thenfluw-betvteenness
is a measure of the number of paths that pass through a possum along the shortest paths bet\veen all other possums (Freeman et al . , 1 99 1 ). It could be envi saged an indiYid ual that lies on the shortest pa
th regulates the flowof
com
mu nicatio ns betweentYvo indirectly linked indiv1 duals (Borgatti, l 995). The p rominence or imp
o
rtance of each p ossum i n the netv\ o rk is the number of direct and indirect connections pass
ing through it as a proportion o f thetotal flow i
n the network. Thi s normal ises the score and allows comparisons between d i lTerent networks (EYerett and Borgatti, 1 999). Flovv betweennesswhere m jk (x i) is the maximum flow that passes from x 1 to x k through possums x 1 along the shortest paths and where j < k and i *
j * k (Freeman et al. ,
1 991 ).The higher the flow betweenness score the more prominent and i nfluential is t he possum I f an individ ual with high flow-betweenness centrality is removed from the network, the speed and certainty of transm ission from a random indi vidual within the network to another is more affected
t
han i f an indivi dual with a low score is removed (Borgatti, 1 995).The .flow-betweenness centralisation index is an overall measure of the variability in the network. It i s the sum of the indi vidual scores divided by the total possible score for the group. A minimum of zero arises when all individuals have exactly the same flow betweenness index and a maximum of one when one individual lies on the links between al l others (Wasserman and Faust, 1 998). In the context of disease transmission, assuming that contact is i mportant, infection wil l spread quickly in a network with high
tlO\v
betweenness index because, even i f an animal with low flow-betweenness score is infected first, it should not take long for another with high flow-betweenness score to become infected, and from then on infection s hould spread very quickly.
The social i nteraction that was measured in this study was den sharing. A possum sharing a den w!th another possum was considered a pair and each pair forming was termed an ' interaction · . Each o ther i ndivid ual a possum shared with during an observat ion period was termed a ' partner', which was recorded as a dichotomous variable (0 = no
recorded pairing, 1 = contact). A social
network
was constructed using the number of partner pairs that formed and the frequency of their interactions. Social network analysis was performed with the SNA software UCINET for Windows version 5 . 1 . 1 . 1 (Analytic Technologies, Harvard, Massachusetts, USA). The relations within the social network w·ere Yisualised as t\;o,io-dimensional graphs generated using the network graphing software KrackPlot (version 3 .2 ; Krackhardt et al . , 1 994).The closeness and flow-betweenness scores for Groups
B,
C and D were compared between periods using the t test. Non-parametric tests w·ere used for Group A because of large differences in the variances between the observation periods. For Group A the closeness and flow betweenness scores were compared between periodsfirstly
with t he1 05
Kruskal-Wall is test and, where significant (p < 0.05), painvise comparisons were made
using the Mann-Whit ney U test. For the M ann-Whitney U test the significance was set at p
<0 0 1 because multiple pairwise comparisons vlere made. Social ranking of possu ms
within al l groups
.
as d etermined by closeness and flow-betweenness scores, w ere compared between observation periods using Spearman ' s rank order correlation.Logistic regression analysis was used to analyse the relationship betvveen the risk of infection and the potential risk factors of vaccination status, group and their interaction terms, as well as the social behavi our scores for closeness and flow-betweenness scores. The analysis was performed using S AS (version 8.0; SAS Institute
.
Cary, North C arol i na).Results
Den use allll den sharing
Box and \..vhi sker plots o f the number of different dens used by an indi vidual possum,
the number of other possums with which each possum shared a den
(part
ners) and thefrequency of den sharing (interactions) are shown in Figures 5 . I A, 5 . 1 8 and 5 . 1 C,
respectively.
]
4 "' § � 3 0 ... ll Ez 2
AI A2 A3 A4 81 82 Cl C2 01 02Group and Period
A1 A2 A3 A4 81 82 C1 C2 01 02 Group and Period
Figu re. 5. 1 B 10 2 A1 Figu re. 5.1 C A4 C1 C2 01 02 Group and Period
Figure. 5.1. Obsenations on the behaviour of four groups (A, B, C and D ) of
communally housed captive brushtail possums. Each gt·oup was observed between 2 and 4 times. For Group A the environment of the pen was changed between pet·iods
Al and A2, A2 and A3, and A3 and A4, and for Group B, 3 new members wet·e added
to the gt·oup between Periods B l and B 2.
A. The n um ber of dens used per 7 days for each of the four groups (A, B, C and D) at
1 07
B. The number of othe•· possums in the g•·oup that each possum shared dens with ( partners) pe•· 7 days.
C. The number of interactions each possum had with other members of the group
(interactions) pe•· 7 days.
Den use: There was considerable \·ariation in the \·vay indi vidual possums used the
availabl e dens (Figure 5 . 1 A). The mean number of dens used by al l p ossums across all
obserration periods was 2.9 ( range 0 to 4.7) per 7 days. In Group A there vvas a sign i ficant d ecline in the numb er of d ens used in Periods A l , A2 and A3 compared to Period A4 (p <
0. 05 ). In Group B there was a s ignificant dec li ne in the number of dens used in the second period (p < 0. 05). There was also an increase in the number of dens used in Group C,
how erer this \vas not statistical ly signi ficant (p=0.46). In Group D there was a signi ficant increase in the number of dens used. ln Period D l , one possum did not use any dens but s l ept on the ground or on elevated
\·Val kways.
Partners: There was considerable variation i n the number of other possums with ,;�,,hich
each possum shared a den. The median for al l possums oYer all peri ods was 2
(r
ange 0 to1 0 ) per 7 days
.
The medi an number of partners decreased i n all groups bet\Yeen the first and l ast obseryation periods (Figure 5 . 1B).
There was a sign ificant difference between obserYation periods for Groups B (p < 0.00 1),
C (p < 0.00 1 ) and D (p < 0.00 I ). In Group A there was no differences in the mean number of partners per possum for Periods A I , A2 and A3, however the val ues were signi fi cant ly lower in Period A4 compared with each ofthe other three periods ).
Interactions: There was considerable variation m the frequency of den-sharing
interactions among possums and the median decreased in al l groups in successive observation p eriods (Figure 5 . 1 C). The medi an for all possums over all periods was 3 . 5 (range 0 to 23). There were s igni ficant differences between the first and second periods for Groups
B
(p<O.OO l ), C (p<O. OO I ) and D (p<O.OO I ). ln Group A there were significant decreases between Periods A l , A2 and A3 when compared with A4 (p<O.OO l for each comparison) and when Period A2 was compared to Period A3 (p = 0.029).Social uetwork analysis
Box and
whiskerplots
for closeness and flow betweennessfor each period are sho'vvn in
F igures
5 . 2
and 5 . 3 respectively. Closeness centralisation indices andflow
betweennesscentral isation
indices
foreach period are shown i n Table 5 . 2.
Table
5.2.
The effects of time and changes to their· environment on the structure of the social networ·k of communally housed captive possums: Changes in closeness and flow betweenness
Group
P
er
iod
Centralization
inde
xCloseness
Flow-betweenness
A
A l 8 1 0 A2 1 6 8 A3
7
1 0 A4 "2 1
;)B
B l
1 6 8B2
532
cC l
29
4
C2
7
1 3
D
D l
2 7
4
D2
61 4
Closeness: In