smoking behaviour and friendship

1
some results from Scottish data
Topic smoking behaviour and friendship Problem
influence and/or selection Theory drifting
smoke rings (Pearson, West, Michell) Data
three wave panel 959697, school year group,
age 13-16 Method SIENA modelling
2
some results from Scottish data

• skip through to Theory
• Smoke rings (Pearson Michell 2000) and
• Drifting smoke rings (Pearson West 2003)
• Group position determines smoking
  behaviour (influence effects)
• peripherals most unstable? (PM)
• peer pressure strongest in groups? (OD)
• isolates most stable smokers? (EB)
• Smoking behaviour determines group
  position (selection effects)
• peripheral smoking rewarded by acceptance
• in a smoking group? (PW)
• group smoking punished by rejection in a
• non-smoking group? (OD)
• isolate smoking breeds further isolation? (PW)

3
some results from Scottish data
Modelling (A) group position as local cohesion,
e.g., reciprocity
group member is embedded in many reciprocal
dyads
isolate is not part of any reciprocal dyad
peripheral is attached to others, but not
reciprocated
Problem reduction to one explanatory dimension
4
some results from Scottish data

• Modelling
• (A) include effects of
• reciprocity
• similarity
• similarity reciprocity
• into the objective functions for network
• and/or behavioural change of actor i.

5
some results from Scottish data

• Modelling
• (A) include effects of
• reciprocity
• similarity
• similarity reciprocity
• into the objective functions for network
• and/or behavioural change of actor i.

similarity
6
some results from Scottish data
Modelling (B) group position as specific
configuration of the neighbourhood
group member belongs to clique of three
isolate has no incoming ties
peripheral is unilaterally attached to group
Problem reduction of explanatory data , loss
of statistical power
7
some results from Scottish data

• Modelling
• (B) include effects of
• isolate status
• group status
• peripheral status
• similarity
• group status similarity
• peripheral status similarity
• into the objective functions for network
• and/or behavioural change of actor i.

8
some results from Scottish data
Results (A) SIENA target statistic
descriptives Average statistic observed per
actor and time interval. Second column
expectations under independence.
9
some results from Scottish data
Results (A) SIENA parameter estimates basis
model
10
some results from Scottish data
Results (A) SIENA parameter estimates basis
model
11
some results from Scottish data
Results (A) SIENA parameter estimates basis
model
12
some results from Scottish data
Results (A) SIENA parameter estimates basis
model
13
some results from Scottish data
Results (A) SIENA estimates extended models
similarity reciprocity in network
model (all other parameters barely
change)
14
some results from Scottish data
Results (A) SIENA estimates extended models
similarity reciprocity in behavioural model
15
some results from Scottish data
Results (A) SIENA estimates extended models
same model with prev. estimates as starting
value
16
some results from Scottish data
Results (B) SIENA target statistic
descriptives Average statistic observed
per actor and time interval.
17
some results from Scottish data
Results (B) SIENA parameter estimates group
position in behavioural model
18
some results from Scottish data
Results (B) SIENA parameter estimates same
without peripheral status effect
19
some results from Scottish data

• Influence question
• Are peripherals most unstable?
• Yes
• influence is strongest in asymmetric
  relationships
• but
• results do not strictly distinguish between
  peripherals and isolates.
• If a peripheral is attached to a homogeneous
  group, the asymmetric influence effect can be
  decisive.

20
some results from Scottish data

• Influence question
• Is peer pressure strongest in groups?
• Definitely not.
• reciprocity cools down the similarity effect,
  influence is strongest in asymmetric
  relationshipss
• (effect not shown here) there is a
  non-significant preference for group homogeneity
  after controlling for reciprocated similarity
• There is peer influence, but not predominantly in
  groups.

21
some results from Scottish data

• Influence question
• Are isolates most stable smokers?
• No.
• neither isolation nor indegree have a significant
  impact on behavioural preference.
• On the contrary
• isolates may be extremely unpredictable (as there
  is no reciprocation of friendship tempering the
  influence of their perceived friends).

22
some results from Scottish data

• Selection question
• Is peripheral smoking rewarded
• by acceptance in a smoking group?
• Not this easily.
• main effect simrec on network is negative, so
  there is a net negative tendency to have
  additional smokers in a smoking group
• question is does tie creation differ here from
  tie dissolution?

23
some results from Scottish data

• Selection question
• Is group smoking punished
• by rejection in a non-smoking group?
• Not either.
• main effect simrec on network is negative, so
  there is a net positive tendency to have smokers
  in a non-smoking group.
• question is again does tie creation differ here
  from tie dissolution?

24
some results from Scottish data

• Selection question
• Does isolate smoking breed further isolation?
• Indirect evidence supports claim
• smokers are significantly less chosen as friends,
• (effect not shown here) unpopular others are
  significantly less chosen as friends isolates
  are extreme unpopulars

25
some results from Scottish data
Literature EB S. Ennett K. Bauman (1993).
Peer Group Structure and Adolescent Cigarette
Smoking A Social Network Analysis. Journal of
Health and Social Behavior 34(3) 226-36. OD E.
Oetting and J. Donnermeyer (1998). Primary
Socialization Theory the Etiology of Drug Use
and Deviance. Substance Use and Misuse 33(4)
995-1026. PM M. Pearson L. Michell (2000).
Smoke Rings Social Network Analysis of
Friendship Groups, Smoking, and Drug-Taking.
Drugs Education, Prevention and Policy 7(1)
21-37. PW M. Pearson P. West (2003).
Drifting Smoke Rings Social Network Analysis
and Markov Processes in a Longitudinal Study of
Friendship Groups and Risk-Taking. Connections
25(2)59-76.