Professional Documents
Culture Documents
To cite this article: Naomi Smith & Tim Graham (2017): Mapping the anti-vaccination movement
on Facebook, Information, Communication & Society, DOI: 10.1080/1369118X.2017.1418406
Introduction
This paper examines the structure and discourse of anti-vaccination public Facebook
pages and considers how the properties of anti-vaccination networks on Facebook may
be analogous to social movements. Understanding pockets of resistance to vaccination
as a public health exercise provides important insights into how these attitudes may be
effectively countered. Effective disease prevention is contingent on high levels of vacci-
nation compliance and coverage within networked populations. When these networks
of coverage are disrupted or begin to disintegrate, diseases such as pertussis (or whooping
cough), measles, and polio re-emerge. Globally, there has been an increased politico-legal
mandate on maintaining high levels of vaccination amongst the community.1
Efforts to increase community vaccination rates to effective standards have not been
without resistance. There have been concerted attempts, both online and offline, to pre-
vent the passage of legislation via techniques of intimidation and coercion, such as
CONTACT Naomi Smith n.smith@federation.edu.au School of Arts, Humanities and Social Sciences, Federation
University Australia, PO Box 3191, Gippsland, VIC 3841, Australia
© 2017 Informa UK Limited, trading as Taylor & Francis Group
2 N. SMITH AND T. GRAHAM
‘doxing’,2 stalking, and death threats (Diresta & Lotan, 2015). This suggests that individ-
uals who are opposed to vaccination, known as anti-vaccinators or ‘anti-vaxxers’, are
serious about protecting their refusal to vaccinate, and use a variety of means to mobilise
the cause. Anti-vaccinators oppose all forms of childhood vaccinations and believe that
vaccinations are toxic, and cause a variety of illnesses and adverse reactions including
autism. As most anti-vaccination communication and organisation takes place online
(Diresta & Lotan, 2015), public social media sites such as Facebook and Twitter provide
rich sources of information regarding the dynamics, discourse characteristics, and net-
worked dimensions of the anti-vaccination movement, making social media a significant
sight for further scholarly analysis.
research that does exist has examined the impact of ‘Web 2.0’ in vaccination decision mak-
ing arguing that users of Web 2.0 applications (social media) have the capacity to influence
and shape public discourse regarding vaccination in a viral manner both positively and
negatively (Betsch et al., 2012). This is in part due to the networked properties of social
media like Facebook. As such, discourse needs to be considered in conjunction with network
properties that make the anti-vaccination community possible online.
Social media has also fuelled the growth of ‘consumer autonomy’ in health care, where
patients are expected and have been encouraged to take an active role in managing their
own health (Dutta-Bergman, 2004). However, the growth of patient ‘consumer autonomy’
has also created ‘flattening’ of expertise, where power shifts from doctors to patients. The
redefinition of expertise has created an environment for anti-vaccination activists to effec-
tively spread their message (Kata, 2012). Insofar as vaccination can be considered as a
‘health problem’, the emergence of public anti-vaccination pages on Facebook can be
Downloaded by [University of Florida] at 04:06 27 December 2017
understood as compelling sites of social support for those who refuse to vaccinate. How-
ever, presently little is known about these public Facebook page communities, including
foundational knowledge about the size of these communities, who participates, how
users are connected within the space, and what kinds of topics are discussed via user com-
ments. This is a gap we aim to addresses in the analysis that follows. As Smith and O’Mal-
ley (2016) argue, online networks like Facebook, together with portable digital devices, ‘act
as a catalyst to mobilize individual and collective protest that … serves both personal and
communal ends’ (p. 5). In this paper, we make the significant step of applying methods
from computational social science towards understanding the anti-vaccination movement
on Facebook as a site of mobilisation for anti-vaccination practices. In order to address
this, we are focussed on determining the networked properties of the anti-vaccination
communities on Facebook selected for analysis. This includes accounting for their size,
shape, and levels of connectedness. Additionally, we are also interested in the topics of dis-
cussion, or the discourse on these pages, as previous research by Davies et al. (2002) and
Kata (2012) suggest that these are potent motivators. However, a large-scale analysis of
anti-vaccination is presently lacking from scholarly literature.
and public health are frequently reported in the news. While the anti-vaccination commu-
nity is still comparatively small, it manages to garner a lot of attention and most of this
attention is a result of content produced in online communities loosely structured around
social media sites such as Facebook, Twitter, and YouTube. Comparing data on anti-vac-
Downloaded by [University of Florida] at 04:06 27 December 2017
cination community across these three platforms is outside the scope of this research.
We have chosen to focus on Facebook as it is still the most popular social network site
in the world and has the broadest user base. A purposive sample of six Facebook pages was
selected by triangulating data (Denzin, 1970) from both Australia and North America,
providing two important, albeit considerably different, sites of current anti-vaccination
activity (see Table 1). Purposive sampling also helped us identify sites that were relevant
to the anti-vaccination movement, and provided interesting and important insights on the
anti-vaccination movement. In addition, conducting large-scale data analysis (detailed
below) the sites were also reviewed before data collection to ensure the timeline for
data collection (14 April 2013 and 14 April 2016) would be analytically useful (Smith &
O’Malley, 2016).
These sites including both ‘community’ pages and pages run by a public figure (e.g.,
Dr Tenpenny). Pages concerning a public figure has more like than community-based
pages, and Australia pages had fewer ‘Likes’ than pages that were more focus on North
America in tone and content. This is not surprising, given the relative population sizes of
Australia and the North America. The research sites we selected were chosen using the fol-
lowing criteria. The Facebook pages had to be explicitly anti-vaccination in focus, available
to the public, and easily discoverable through keyword searches such as ‘vaccine concern’,
‘vaccine choice’, ‘vaccines and autism’, and ‘anti-vaccination’. In addition, Facebook pages
must be frequently updated, that is, daily or with updates every other day. Pages must
also have a minimum of 1000 ‘likes’, as well as evidence of user activity and interaction,
including double digit likes on posts, as well as comments and ‘sharing’ of posts.
These criteria helped ensure we had active sites for data collection and a large enough
sample size to elicit sufficiently generalisable findings regarding the structure and compo-
sition of the anti-vaccination movement on Facebook. A range of methods was used to
answer the research questions of this study:
These research questions broadly fit into two approaches: social network analysis and text
analysis. This section provides details about the methods used for each of these approaches.
INFORMATION, COMMUNICATION & SOCIETY 5
network, an undirected edge is created if the users both co-commented on any post. Thus we
set a very low bar for creating a relationship between users: two users are connected merely if
they have both commented on the same post within the 3-year period for a given page-level
network. Additionally, each edge in the ‘user comments’ networks has a weight value that
represents how many times two users co-commented on any posts together. This is illus-
trated in Figure 3 by the ‘thickness’ of each edge.
INFORMATION, COMMUNICATION & SOCIETY 7
Figure 3. Structure of the Facebook one-mode network projection of user co-comment activity.
8 N. SMITH AND T. GRAHAM
Table 3. Node and edge count statistics for the page-level networks.
Page/network Nodes Edges
AVN living wisdom 12,978 95,392
Rage against vaccines 20,071 175,540
Age of autism 13,136 85,699
GMAOQV 13,555 76,844
No vaccines Australia 8693 63,186
Vaccine info 215,473 2,326,714
Movement-level network 257,549 2,823,375
within each page derives from ‘transient’ users who only post once or twice within the
3-year period.
The next step of the analysis examined what percentage of users were active between
multiple anti-vaccination pages. Are users active on one page or do they participate across
Downloaded by [University of Florida] at 04:06 27 December 2017
multiple pages? For each page-level network i and j, we calculated how many users who
participated in network i also participated in network j, expressed as a percentage.
Table 5 shows the percentage of users who participated between pairs of page-level net-
work. On average, only about 11% of users were active across any two given pages. How-
ever, an exception was the ‘Vaccine Info’ page network, whereby approximately one-third
of users were also active on other pages. Whilst this makes sense given the large size of the
‘Vaccine Info’ page, it also indicates that this page is singularly important within the anti-
vaccination movement on Facebook.
We now turn attention to the one-mode ‘user co-comment’ networks that were pro-
jected from the page-level two-mode networks. As discussed previously, these networks
depict only users (nodes) and comment activity (edges), whereby an edge between two
users means that they both commented at least once on any post within a given network
over the 3-year period. These one-mode projections of the networks provide a different
picture of the anti-vaccination movement on Facebook, because the focus is on users
and their comment activity (rather than ‘likes’). As argued previously, commenting
requires more effort and is more involved than simply ‘liking’ a post. Comments also con-
tribute differently to the spread of anti-vaccination discourse and ideas, given that users
are able to read each other’s’ comments, interpret and learn from them, and engage in dis-
course by posting their own comments. Therefore, these networks provide specifically
interesting perspectives on user (co)participation and discourse dynamics within each
page and across the entire movement-level network.
Table 6 provides descriptive statistics of each projected user co-comment network. As
the ‘edges’ values in Table 5 shows, a large number of users co-comment together within
each network during the 3 years. Indeed, across the entire movement-level network there
are nearly 17 million occurrences of users co-commenting on a post. However, the ‘den-
sity’ values of the networks reveal that of all the possible co-comments between users, only
a small percentage are actually observed. In other words, across the entire movement-level
Downloaded by [University of Florida] at 04:06 27 December 2017
network, about 2% of users have co-commented with another user on any post during the
3-year period. This suggests that, on the whole, users are not acting together in a coordi-
nated fashion, but instead supports the idea of users as ‘transient’ visitors, as the previous
analysis indicated. However, even if two users have not directly co-commented together
on the same post, this does not mean that they are not potentially influenced by each
other’s comments or connected in analytically interesting ways.
To examine this further, we assess whether the user co-comment networks exhibit the
property of being ‘small world’ networks (Watts & Strogatz, 1998). As Shirky argues, small
world networks ‘have two characteristics that, when balanced properly, let messages move
through the network effectively’ (2008, p. 215). The first characteristic is that small groups
of nodes in the network are highly clustered and inter-connected. The second characteristic
is that large groups of nodes in the network are sparsely inter-connected. Small world net-
works enable infectious diseases to spread much more quickly and easily than other types
of networks, such that the dynamics of the network is an ‘explicit function of structure’
(Watts & Strogatz, 1998, p. 441). Small world networks are also interesting because
they are robust and resistant to damage. That is, randomly removing nodes from the net-
work will not significantly impact the effectiveness and dynamics of the network. In terms
of the anti-vaccination movement on Facebook, if these networks are ‘small world’ then
this has repercussions for the dynamics of the movement and how ideology is diffused
via comment activity, which is discussed later.
We use the approach of Watts and Strogatz (1998) to assess whether the user co-com-
ment networks are ‘small world’. A network is ‘small world’ if it satisfies two conditions.
First, its average local clustering coefficient must be much greater than a random network
generated from the same set of vertices. Second, the mean shortest path length of the
network must be approximately the same as the associated random network. We con-
structed six random networks corresponding to the vertex sets of the six user co-comment
networks, along with a seventh random network corresponding to the movement-level co-
comment network. The random networks were created using the Erdős-Rényi model
implementation in the ‘igraph’ R package (Csardi & Nepusz, 2006).
The results in Table 6 suggest that each of the networks is ‘small world’ given that they
approximately satisfy the two conditions set out by Watts and Strogatz (1998). In this way,
the average local clustering coefficient scores are much greater for the observed networks
versus the random networks (bolded in brackets), and the mean shortest path length
scores are approximately the same between the observed and random networks.
The generation of networks from the Facebook data resulted in seven networks, that is,
one network for each anti-vaccination page and one large-scale network constructed
from the entire dataset across all six pages. A range of SNA methods were employed to
analyse the network data, including descriptive statistics for network characteristics
(Kolaczyk & Csárdi, 2014, pp. 44–58). In addition to understanding how anti-vaccination
Facebook pages are structured as networks we also wished to examine the gender compo-
sition of the networks, in order to ascertain the extent to which the anti-vaccination move-
ment on Facebook is gendered.
Very little information is provided by the Facebook API regarding the individual
characteristics of users. Currently, only the user names are made available through the
public search API. However, for this study we wanted to understand the gender5 of
users who are participating in the Facebook anti-vaccination networks. To achieve this,
we modelled the gender of users based on their first or ‘given’ name. We used the method
implemented by Mullen (2016) in the ‘gender’ software package. As Mullen (2016)
describes, this approach ‘encodes gender based on names and dates of birth using historical
datasets. By using these datasets instead of lists of male and female names, this package is
able to more accurately guess the gender of a name, and it is able to report the probability
that a name was male or female’ (p. 1). Thus for each user in our networks we extracted the
first name and used this to predict whether the user was more likely to be male or female.
For the ‘gender’ package functionality, we used the ‘SSA’ method, which queries names
based on the U.S. Social Security Administration (SSA) baby name dataset (from 1932
to 2012). This returned a list of the first names along with the probability for male and
female, or an ‘unknown’ value for names that do not appear in the SSA dataset.
As Table 7 shows, the findings suggest that the majority of people who participate
within the anti-vaccination movement on Facebook are female. On average, the ratio of
male to female is approximately 1:3, indicating a pronounced gender imbalance in
terms of participation within the anti-vaccination movement on Facebook. We also
find that the gender imbalance is even more skewed for the most active users, which we
define as the top 10% of users by number of comments and likes. As the figures in brackets
in Table 6 show, men are consistently under-represented within the set of most active
users, whilst women are consistently over-represented. Notably, a small percentage of
users had names that were not able to be modelled using the gender analysis, resulting
in the ‘unknown’ observations.
INFORMATION, COMMUNICATION & SOCIETY 11
Using the node-level ‘gender’ attribute we can also examine the distribution of edges by
gender, that is, the proportion of directed edges that have a female versus male source
Downloaded by [University of Florida] at 04:06 27 December 2017
node. This provides a better indication of user activity by gender, i.e., likes and comments
on the Facebook posts. As Table 8 shows, the edge-level user activity within the networks
is slightly more skewed in terms of gender distribution, with an increase of 4.2% compared
to the node-level gender distribution. Across the entire movement-level network, 21.9% of
user activity is male, 71.4% is female, and 6.7% is not estimated (gender unknown).
Figure 3 provides a network visualisation of the anti-vaccination movement, where the
colour of the nodes represents gender. Male nodes are yellow, female are purple, and
the ‘posts’ nodes are light blue. The network is overwhelmingly purple, reflecting the pre-
dominantly female gender of the users (Figure 4 and Table 9).
A key question for LDA topic modelling is how many topics k to choose. The value of k
must be specified prior to running the algorithm. There are several approaches to selecting
the number of topics, although no standardised practice is currently agreed upon. The
hierarchical Dirichlet process is often regarded as an optimal approach to choosing k,
however, our dataset was too large for it to be tractable (Teh, Jordan, Beal, & Blei,
2006). Therefore, we used a data-driven and mathematical approach combined with a
heuristic approach that aimed to maximise the interpretability and usefulness of the topics.
First, we cleaned the text data to prepare it for analysis. This involved three main steps:
removing extremely infrequent and extremely frequent words using term frequency-
inverse document frequency (tf-idf); converting all words to lowercase; and removing
punctuation. Second, we used Ponweiser’s (2012) approach to model selection by harmo-
nic mean. We ran 35 different LDA models using values of k ranging from 2 to 70
(sequenced by intervals of 2). From these models, we could determine which value of k
produced the maximum harmonic mean of the log-likelihood, given the data. Third, we
plotted the harmonic mean of log-likelihood values by each number of topics k. Using Cat-
tell’s scree test (Cangelosi & Goriely, 2007, p. 10), we were able to identify the inflection
point or ‘elbow’ in the graph, at which there was no significant increases in log-likelihood
values for the models. There were no significant increases after k = 26. Using Cattell’s scree
test as a heuristic guide, we were able to select 26 topics as the number for further analysis
and interpretation.
The final part of analysis used LDA topic modelling to examine what kinds of ‘topics’
might be driving anti-vaccination discourse on Facebook. As discussed earlier, the chosen
number of topics was 26. Out of all the topic only one topic (Topic 8 in Table 10) was
identified as a ‘junk’ topic (Nikolenko, Koltcov, & Koltsova, 2015). The word probabilities
in this topic word were not interpretable and not relevant to our analyses. In addition to
this, some topics also include ‘junk’ terms, words that were not interpretable in relation to
Downloaded by [University of Florida] at 04:06 27 December 2017
others in the topic. These terms are generally considered to be an artefact of social media
data. These terms were manually excluded from the final analysis presented below. The
findings of these results are summarised in this section. Table 10 provides a list of the
topic labels, which were interpreted based on the terms that were assigned with highest
probability to each topic. In order to add interpretive depth to the topic models presented,
we have labelled these topics with a qualitative label that best reflects the topic. These
qualitative labels are based on our observation of anti-vaccination pages over the past
year, which occurred alongside data collection and analysis. The qualitative labels were
further informed by existing anti-vaccination literature.
may reinforce and cement anti-vaccination beliefs. Our data also suggest that participants
are moderately active across several anti-vaccination Facebook pages, suggesting that
users’ activity on anti-vaccination is more than just a product of Facebook’s recommender
system. Liking and actively commenting on a number of anti-vaccination pages across
Facebook suggests that those who are involved in this network may develop a pattern
of activity and involvement across multiple pages, creating a filter bubble effect that
reinforces anti-vaccination sentiment and practice. However, it is difficult to discern
from the data to what extent the filter bubble is created through users’ own agency and
activity, and how much is influenced by the algorithmic structure of Facebook, whereby
Facebook actively targets users with content they would be more likely to click on and
relate to. As discussed later in this paper, further research is needed to fully address this
question, and understand how health information and communities on social media
both shape, and are shaped by, day-to-day practice.
Further to this, the network structure of anti-vaccination Facebook pages has impli-
cations when considering how ‘robust’ or open to change this social movement might
be. As detailed previously, anti-vaccination Facebook pages exhibit the signature charac-
teristics of a ‘small world’ network structure. In small world networks, information dif-
fuses quickly and easily through the network, in this instance through user-generated
comments. As with other aspects of this analysis, we argue that it is difficult to say whether
the ‘small world’ characteristics of the networks examined in this paper are due to the
nature of the anti-vaccination movement itself, or are an artefact of Facebook as a plat-
form. Both outcomes are equally interesting. The former suggests that social movements
(like anti-vaccination) may inevitably develop as ‘small world’ networks structure that is
further amplified and made visible online. If it is the latter, this demonstrates that Face-
book as a platform has important implications for the dynamics, spread, and durability
of social movements outside of the specific case examined here. Indeed, if the materiality
or architecture of Facebook shapes networks towards ‘small-worldness’, this suggests that
such platforms may be instrumental for the anti-vaccination movement and social move-
ments more broadly to blossom, flourish, and resist being dismantled or disrupted by
outside influences.
The large numbers of ‘likes’ and posts on the anti-vaccination sites examined suggest a
popular and active community. In addition to like, and commenting on a post, posts are
INFORMATION, COMMUNICATION & SOCIETY 15
also shared widely as detailed previously in Table 2. This finding suggests that the anti-vac-
cination community has scope and reach outside of those who immediately engage with
the public Facebook pages examined in this study. For the period of data collection, the
total number of shares across all pages sampled in this research was 2,024,672.6 The num-
ber of shares also far out-paces the number of comments for this same period, suggesting
that sharing is a more common mode of participation than commenting or liking posts.
Sharing also suggests that the anti-vaccination network extends further than is apparent
from the Facebook pages analysed in this research. However, the research conducted
here is limited by the combined public/private nature of Facebook. While public Facebook
pages do provide a wealth of network information, we are unable to gather information
about how information shared from anti-vaccination pages disseminates through private
Facebook pages or personal social media networks. Nonetheless, the data suggests that
sharing may be an important function in spreading anti-vaccination information and
Downloaded by [University of Florida] at 04:06 27 December 2017
10 all appear to accord with conspiracy-style beliefs in which the government and media
are key actors in underplaying, denying, or perpetuating the perceived harms caused by
vaccinations. These include: media cover-up or denial of the extent of vaccination injury
and death (Topic 3); Bill Gates’ involvement in the spread of Zika virus within Brazil and
beyond its borders (Topic 5); and chemtrails (Topic 10), which is a belief that the vapour
trails emitted by aircraft are chemical compounds sprayed by the government and
designed to subdue the population and/or control the weather (Oliver & Wood, 2014).
The prevalence of conspiracy-style thinking is unsurprising. Survey research under-
taken by the Cooperative Congressional Election Studies indicates that 55% of respon-
dents in 2011 agreed with at least two of the conspiracy theories they were presented
with. This suggests that conspiracy-style thinking is relatively prevalent in the general
population, and perhaps particularly pronounced among those who hold anti-vaccination
beliefs (Oliver & Wood, 2014). Sitting alongside these conspiracy-style beliefs are concerns
Downloaded by [University of Florida] at 04:06 27 December 2017
about state-sanctioned harm or interference with citizens’ bodily autonomy through the
administration of vaccination. In particular, Topic 4 demonstrates how anti-vaccination
Facebook pages commonly compare vaccination to the Holocaust, illustrating a strong
sense of persecution within the anti-vaccination communities examined in this study.
While Topic 9 appears to be an ‘outlier’ in relation to the other topics shown previously
in Table 10, it is also suggestive of a distrust of authority. Anti-vaccination communities
tend to advocate natural remedies for maintaining health or curing illness. These include
avoiding fluoridated tap water, avoiding sunscreen, using coconut oil to improve mouth
health, and generally prescribing food as medicine. The emphasis on natural remedies
is in keeping with the anti-vaccination movement’s distrust of the broader medical and
scientific community, which they regard as irrevocably corrupt.
While this research analysed a large volume of empirical data, our findings are limited
by the number of Facebook pages sampled. This study has also focussed on analysing anti-
vaccination on Facebook, limiting our ability to generalise about anti-vaccination activity
across social media platforms. However, it is not the goal of this research to account for all
anti-vaccination content online – although more comprehensive study is warranted – but
to provide a foundational, empirically informed account of the anti-vaccination move-
ment in a social media context. The results of this investigation suggest a robust and highly
gendered network structure that has a strong sense of moral outrage associated with the
practice of vaccination. This ‘righteous indignation’, in combination with the network
characteristics identified in this study, indicates that anti-vaccination communities are
likely to be persistent across time and global in scope as they utilise the affordances of
social media platforms to disseminate anti-vaccination information. Concerns about vac-
cination reveal a community that feels persecuted and is suspicious of mainstream medical
practice and government-sanctioned methods to prevent disease. In a generation that has
rarely seen these diseases first hand, the risk of adverse reaction seems more immediate
and pressing than disease prevention (Davies et al., 2002).
Notes
1. Countries such as France issue fines for non-compliance. Similar legislation removing per-
sonal belief exemption has also been passed in California and Australia (Klapdor & Grove,
2015).
INFORMATION, COMMUNICATION & SOCIETY 17
Disclosure statement
Downloaded by [University of Florida] at 04:06 27 December 2017
Notes on contributors
Naomi Smith is a Lecturer in Sociology in the School of Arts, Humanities and Social Sciences at
Federation University Australia (Gippsland). She received her PhD in Sociology of The University
of Queensland. Her current research projects examine intersection of health and wellness practices,
physical spaces and social media [email: n.smith@federation.edu.au].
Tim Graham is Postdoctoral Research Fellow at the Australian National University, Australia, with
a joint appointment in the Research School of Social Science and the Research School of Computer
Science. His research combines social theory and computationally-intensive approaches to analys-
ing, understanding, and predicting social phenomena. The holds a PhD in Sociology from the Uni-
versity of Queensland [email: Timothy.Graham@anu.edu.au].
References
Betsch, C., Bohm, R., & Chapman, G. B. (2015). Using behavioral insights to increase vaccination
policy effectiveness. Policy Insights from the Behavioral and Brain Sciences, 2(1), 61–73.
Betsch, C., Brewer, N. T., Brocard, P., Davies, P., Gaissmaier, W., & Haase, N. (2012). Opportunities
and challenges of Web 2.0 for vaccination decisions. Vaccine, 30(25), 3727–3733.
Blei, D. M., Ng, A. Y., & Jordan, M. I. (2003). Latent dirichlet allocation. Journal of Machine
Learning Research, 3(2003), 993–1022.
Butler, R., & MacDonald, N. E. (2015). Diagnosing the determinants of vaccine hesitancy in
specific subgroups: The guide to tailoring immunization programmes (TIP). Vaccine, 33(34),
4176–4179.
Camerini, L., Diviani, N., & Tardini, S. (2010). Health virtual communities: Is the self lost in the net?
Social Semiotics, 20(1), 87–102.
Cangelosi, R., & Goriely, A. (2007). Component retention in principal component analysis with
application to cDNA microarray data. Biology Direct, 2(1), 2–12.
Csardi, G., & Nepusz, T. (2006). The igraph software package for complex network research.
InterJournal, Complex Systems, 1695. http://interjournal.org/manuscript_abstract.php?
361100992
Davies, P., Chapman, S., & Leask, J. (2002). Antivaccination activists on the world wide web.
Archives of Disease in Childhood, 87(1), 22–25.
Davies, P., Chapman, S., & Leask, J. (2002). Anti-vaccination activists on the world wide web.
Archives of Disease in Childhood, 87(1), 22–25.
Denzin, N. (1970). The research act in sociology. Chicago: Aldine.
18 N. SMITH AND T. GRAHAM
Diresta, R. and Lotan, G. (2015). Anti-vaxxers are using twitter to manipulate a vaccine bill.
Retrieved from https://cran.rproject.org/web/packages/SocialMediaLab/SocialMediaLab.pdf
Durbach, N. (2005). Bodily matters: The anti-vaccination movement in England, 1853-1907.
Durham: Duke University Press.
Graham, T. & Ackland, R. (2016). SocialMediaLab: Tools for collecting social media data and gen-
erating networks for analysis. CRAN (The Comprehensive R Archive Network). Retrieved from
https://cran.rproject.org/web/packages/SocialMediaLab/SocialMediaLab.pdf
Grant, L., Hausman, B. L., Cashion, M., Lucchesi, N., Patel, K., & Roberts, J. (2015). Vaccination
persuasion online: A qualitative study of two provaccine and two vaccine-skeptical websites.
Journal of Medical Internet Research, 17(5), e133.
Gray, N. J., Klein, J. D., Noyce, P. R., Sesselberg, T. S., & Cantrill, J. A. (2005). Health information-
seeking behaviour inadolescence: the place of the internet. Social Science & Medicine, 60(7),
1467–1478.
Kata, A. (2010). A postmodern Pandora’s Box: Anti-vaccination misinformation on the internet.
Vaccine, 28(7), 1709–1716.
Downloaded by [University of Florida] at 04:06 27 December 2017
Kata, A. (2012). Anti-vaccine activists, Web 2.0, and the postmodern paradigm: An overview of tac-
tics and tropes used online by the anti-vaccination movement. Vaccine, 30(25), 3778–3789.
Klapdor, M., & Grove, A. (2015). Budget review 2015–2016 index: No Jab No Pay and other immu-
nisation measures. Parliament of Australia. Retrieved from http://www.aph.gov.au/about_
parliament/parliamentary_departments/parliamentary_library/pubs/rp/budgetreview201516/
vaccination
Kolaczyk, E. D., & Csárdi, G. (2014). Statistical analysis of network data with R. New York: Springer.
Medved, C. E. (2016). Stay-at-home fathering as a feminist opportunity: Perpetuating, resisting, and
transforming gender relations of caring and earning. Journal of Family Communication, 16(1),
16–31.
Mullen, L. 2016. Gender: Predict gender from names using historical data [R package version 0.5.2].
Retrieved from https://github.com/ropensci/gender
Nikolenko, S. I., Koltcov, S., & Koltsova, O. (2015). Topic modelling for qualitative studies. Journal
of Information Science. Advance online publication. http://jis.sagepub.com/content/early/2015/
12/05/0165551515617393.full
Oh, H. J., Lauckner, C., Boehmer, J., Fewins-Bliss, R., & Li, K. (2013). Facebooking for health: An
examination into the solicitation and effects of health-related social support on social networking
sites. Computers in Human Behavior, 29(5), 2072–2080.
Oliver, J. E., & Wood, T. J. (2014). Conspiracy theories and the paranoid style(s) of mass opinion.
American Journal of Political Science, 58(4), 952–966.
Percheski, C., & Hargittai, E. (2011). Health information-seeking in the digital age. Journal of
American College Health: J of ACH, 59(5), 379–386.
Ponweiser, M. (2012). Latent dirichlet allocation in R (Diploma thesis). Institute for Statistics and
Mathematics, WU (Wirtschaftsuniversität Wien), Austria. Retrieved from http://epub.wu.ac.at/
3558/1/main.pdf
Shirky, C. (2008). Here comes everybody: The power of organizing without organizations. New York:
Penguin Press.
Smith, G. J. D., & O’Malley, P. (2016). Driving politics: Data-driven governance and resistance. The
British Journal of Criminology, 56(1), 1–24.
Smith, N., Wickes, R., & Underwood, M. (2015). Managing a marginalised identity in pro-anorexia
and fat acceptance cybercommunities. Journal of Sociology, 51(4), 950–967.
Teh, Y. W., Jordan, M. I., Beal, M. J., & Blei, D. M. (2006). Hierarchical dirichlet processes. Journal
of the American Statistical Association, 101(476), 1566–1581.
Vaterlaus, J. M., Patten, E. V., Roche, C., & Young, J. A. (2015). Gettinghealthy: The perceived influ-
ence of socialmedia on young adult health behaviors. Computers in Human Behavior, 45, 151–
157. doi:10.1016/j.chb.2014.12.013
Watts, D. J., & Strogatz, S. H. (1998). Collective dynamics of ‘small-world’ networks. Nature, 393
(6684), 440–442.