TheGene'cHeritabilityofSocialand - - PowerPoint PPT Presentation

The
Gene'c
Heritability
of
Social
and
 Poli'cal
Traits:
Introduc'on
to
Twin
Studies


Levi
Li=vay
 Research
and
Methods
Symposium
–
WSU
 January
28,
2010


What
is
the
Central
 European
University?


• In
Budapest,
Hungary
(since
1991)

• US
(and
now
also
EU)
accredita'on

• Working
language:
English
(strictly)

• Only
social
sciences

• Only
grad
school
(MA,
PhD,
LLM)

• One
year
MA
in
Poli'cal
Science

• Diverse
Student
Body

• Big
fat
endowment
by
George
Soros



(read:
Lots
of
Scholarships)


• Send
us
good
students!


Step
#1:
AdmiYng
You
Have
a
Problem


Plutzer
(APSR
2002


Predic'ng
 turnout
 What
do
we
 see
here?
 (beyond
chaos?)
 33
Independent
 Variables
 R2
=
0.32


Two
Dogmas
of
Poli'cal
Science


• Behaviorist
Approach


– Because
internal
mental
processes
and
personal
 preferences
cannot
be
observed,
we
will
focus


• ur
inquiry
on
observed
behaviors
and
revealed


preferences.


• Ra'onal
Choice
Theory


– Human
behavior
is
best
explained
by
assuming
 that
individuals
are
ra'onally
trying
to
maximize
 their
individual
u'lity.


This
Slide
and
the
Previous
Slide
“Stolen”
from
Darren
Schreiber


Model
of
Individual
Behavior:
Ra'onal
 U'lity
Maximiza'on


• Maximiza'on:
we
want
MORE!
(preferences)


• Expected
behavior
can
be
modeled.

• What
Is
U'lity?

• Hard









vs.









Sod























Ra'onal
Choice


2002
Nobel
Prize
Winners:
For
Work
in
 Experimental
Economics


Daniel
Kahneman
 Vernon
Smith


Solu'ons:
“as
if”


• People
might
not
act
ra'onally
but…



Models
work
in
the
aggregate
“as
if…


• Not
interested
in
“as
if”
aggregate
models


Solu'on:
“U'lity
Is
Not
Money”



Pres'ge

Friendship

Being
Nice
Love

Sex

 Power

Goods

Services

Fairness
Happiness

 Gold

Jewelry

Shopping

Accomplishment

 Family

Health

Food

Clothes

Pets

Kids

 Pizza

Beer

Wine
Some'mes
This




 Some'mes
That

Alcohol

Coca~Cola

Music

 Yeah

etc.


TAUTOLOGY



Search
for
preferences


• Need
a
new
theory
to
understand
and


predict
preference
structures.
 

 
 



or


• Need
a
new
theory
of
behavior.

• Unselfish
Behavior
(Altruism)
Literature:


– Evolu'onary
Explana'ons


APSR:
May
2005
Cover
Ar'cle



“…gene'cs
play
an
 important
role
in
shaping
 poli*cal
a-tudes
and
 ideologies
but
a
more
 modest
role
in
forming
party
 iden*fica*on…”


Twin Data

• Most informative family data is twin data
• Two types of Twins

– Monozygotic (MZ) – Dizygotic (DZ)

• Four Sources of Information

– MZ Twins reared Together – DZ Twins reared Together – MZ Twins reared Apart – DZ Twins reared Apart (Latter two are almost non-existent today)

Monozygotic Twins

• Look identical
• From same egg
• Share 100% of

their genes.

• Natural clones
• Reared together

they share some environment

• Same age

Dizygotic Twins

• Share 50% of their genes on average
• Don’t necessarily look exactly the same
• Reared together they share some environment
• Same age (otherwise just like other siblings)

Analytical Approaches

• Correlations (pre-1970) but good heuristic
• Correlate the scores of co-twins across families

separately for MZs and DZs. (rMZ and rDZ)

• What to look for:

– if rMZ > rDZ (MZ twins are more similar to each other): Trait is probably heritable – if rMZ = rDZ: Trait is probably environmental – If rMZ > 2*rDz: Dominance (never seen for social traits)

• Anova/Ancova/Manova/Mancova
(1970‐1977)


– Maybe
s'll
used
with
very
small
samples


• Structural
Equa'on
Models
with
ML
(up
to
date)


Expecta'ons
(More
Formally)

A (Additive Genetic Influence): 2(rMZ – rDZ) (Assumes no Dominance, Gene-Gene or Gene- Environment Interactions) C (Common Environment): 2rDZ – rMZ D (Domiannce): 2rMZ – 4rDZ E (Unique Environment): 1 – MZ (This also includes all sources of deviation from perfect correlation… like measurement error)

Other
Things
to
Watch
Out
For


• Who
is
an
MZ
and
who
is
a
DZ


– Self
report
is
95%
accurate.
(Is
that
good
enough?)


• Mixture
Model
Correc'on
(lets
not
get
too
far
ahead)


– Mul'
Ques'on
Latent
Class
Analysis
 – Genotyping


• No
influen'al
outliers

• No
differences
in
means
and
variances
between













MZ
and
DZ
twins
(and
between
twin
1
and
twin2)


– For
categorical
data,
same
is
true
for
category
thresholds


• Equal
Environemnt
Assump'on
(more
on
this
later)

• No
alterna've
sources
of
variance


– Self
selec'on
into
certain
environments
 – Random
ma'ng
of
parents


The
Univariate
Structural
Equa'on
Model


What
else
can
you
do
with
this
model


• Explain
varinace
of
“phenotype”
with
predictors


Before
Decomposing
Residual
Variance
Into
A,
C
&
E


– Age
and
sex
is
standard.

Can
use
others.


• Sex
difference
(include
DZOS)
4,
5,
6
group
model

• Add
Addi'onal
Family
Members
into
the
Model

• Use
a
measurement
(CFA
or
IRT)
model
to
construct


phenotype


• Test
rela'onships
between
phenotypes.



Decompose
the
covariance
into
A,
C
and
E


Biological
Pathway:
Genomics


DNA
 Genes
 Nucleo'des
 Amino
Acids
 Proteins:
 (Genes
code
for
these)
 Neurons
 Brain
 Behavioral
implica'ons
can
be
assessed
at
each
of
these
steps


What
am
I
doing?

(Survey
Research)


• Na'onal
Survey
of
Midlife
Development
in
the


United
States
(MIDUS)


– Wave
I
was
collected
in
1995‐1996
 – Wave
II
collected
10
years
later
(but
lot
of
missing)
 – Only
used
items
present
on
both
waves
 – Large
Na'onal
Representa've
Sample

 – RDD
CATI
+
Mail
Follow
Up
 – Oversample
of
twins
(and
sibs
and
urban)
 – Singleton
sample
n=3091
 – Twins
sample
nMZ=359
/
nDZ=337
pairs


What
is
Survey
Response
Style?
 Acquiescence
and
Extreme
Response


Cholesky
longitudinal
model
(for
twin
1)


Panel
Cholesky
ACE
Model
Results


Variance and Covariance Decomposed into ACE Acquiescence Total A C E

• Variance 1 0.286** 0.114 0.600***

Covariance 0.569*** 0.286** 0.114 0.169***

• Extreme Resp. Total A C E
• Variance 1 0.196 0.321* 0.483***

Covariance 0.781*** 0.196 0.302+ 0.283***

• +: p<.10, *: p<.05, **: p<.01, ***: p<.001

Personality
(Big
5)


– Extraversion
 – Agreeableness
 – Neuro'cism


Personality A C E

• Agreeableness 0.344*** 0 0.656***

Conscientiousness 0.641*** 0.015 0.344*** Extraversion 0.533*** 0 0.467*** Neuroticism 0.549*** 0.03 0.421*** Openness 0.532** 0.022 0.446***

• **: p<.01, ***: p<.001
• Personality
(Big
5)


– Openness
 – Conscien'ousness


• Work
with
Ma=hew
Hibbing


Personality Acquiescence

• Agreeableness ++

Conscientiousness ?? Extraversion -- Neuroticism + Openness --

• or + : Weak expectations,
• - or ++: Strong expectations

Expecta'ons
(also
on
next
slide)


• Just
Focused
on
Acquiescence


– Personality
does
not
have
a
C
component
 – Extreme
response
does
not
have
a
significant
A


• Extraversion
(‐‐)


– Linked
to
higher
self
confidence
and
risk
taking
 – Leads
to
a
greater
comfort
in
expressing
disagreement


• Openness
(‐‐)


– Also
linked
to
self
confidence
and
risk
taking
(same
as
above)
 – Addi'onally,
there
is
a
known
neg.
rela'onship
between
 cogni've
complexity,
intolerance
of
alterna'ves
and
acq.


• Conscien'ousness
(?)



– Known
pos.
rela'onship
between
organiza'on
and
acq.
 – Conscien'ous
people
are
honest
(ie.
do
not
acquiesce)


• Agreeableness
(++)


– no
need
to
explain
expecta'on


• Neuro'cism
(+)


– Nega've
sense
of
self
efficacy
 – Discoun'ng
of
their
own
opinion

excessive
agreement



• To
simplify
things
we
only
considered
wave
1


– Had
enough
power
to
be
able
to
do
this


• Ran
the
same
Cholesky
decomposi'on
model


– Less
model
restric'ons


Bivariate
ACE
Cholesky
Decomposi'on


Table of Expectations and results (p-values in parentheses) Personality Acquiescence Correlation

• Agreeableness ++ -.049 (.143)

Conscientiousness ?? -.272 (.001) Extraversion -- -.155 (.001) Neuroticism + .374 (.001) Openness -- -.189 (.001) ACE Correlations with Acquiescence A C E

• -- --- ---

Agreeableness: not estimated due to insignificant corr. Conscientiousness: -.240 (.003) .007 (.907) -.039 (.309) Extraversion: -.167 (.010) .010 (.830) .002 (.949) Neuroticism: .210 (.047) .060 (.498) .103 (.004) Openness: -.287 (.002) .041 (.544) .057 (.130)

Results


Problems
with
Linkage
and
GWAS


Candidate
Genes


• SLC6A4
–
Serotonin

Transporter
(a.k.a.
5‐HTT
or
SERT)


– Linked
to
Neuro'cism
(22
cites),
Openness
in
males
(1
cite)
 and
Extraversion
(1
cite)


• DRD4
(and
possibly
other)
Dopamine
Receptors


– Linked
to
Neuro'cism
(4
cites)
and
to
Extraversion
(6
cites).
 Also,
to
novelty
seeking
(common
indicator
of
openness).





First
two
are
available
in
replica'on
dataset
(Add‐Health)


• COMT
–
degrades
catecholamines
(like
dopamine)


– Linked
to
Neuro'cism
(11
cites),
Openness
(1
cite),

 and
Extraversion
(3
cites).



• ISN
–
Insulin


– Linked
to
Conscien'ousness
(1
cite)
and
Neuro'cism
(11
cites)
 – I
see
a
plausible
environmental
explana'on
here


Doing
the
Same
with
Par'cipa'on


• Instead
of
acquescence,
use
poli'cal
par'cipa'on


– Results
corroberate
Fowler
et
al
2008
APSR
well
 – No
wonder
Plutzer
could
only
get
an
R2
of
.32


• Have
results
for
vo'ng
but
also
have
data
on
other
stuff


Univariate A C E

• Obligation to Vote 0.388** 0.002 0.610**
• **: p<.01

ACE Correlations with Acquiescence (Bootstrapped CIs) A E

• -- ---

Agreeableness: .073 (-.12, .154) .089 (-.108, .161) Conscientiousness: .145 ( .053, .231) .062 (-.013, .138) Extraversion: .125 ( .044, .194) .078 ( .015, .147) Neuroticism: .133 ( .052, .211) -.003 (-.077, .072) Openness: -.167 (-.258, .215) -.058 (-.109, .126)

We
also
have
data
on


• How
much
obliga'on
would
you
feel
to
vote
for
a


law
that
would
help
others
worse
off
than
you
but
 would
increase
your
taxes?


• How
many
hours
per
month
do
you
volunteer
work


for
poli'cal
organiza'ons
or
causes?


– We
also
have
other
other
volunteer
ac'vi'es
for
comp


• How
many
dollars
per
month
do
you
controbute
to


poli'cal
organiza'ons
or
causes?



– We
also
have
other
other
organiza'ons
for
comparison


• Also,
there
are
other
psych
variables
we
didn’t
even


look
at
(will
do
probably
for
ISPP,
deadline
is
Feb
6.)


The
Equal
Environment
Assump'on


• No
it,
does
not
mean
that
MZ
and
DZ
twins’


environment
is
equal.
(Common
mispercep'on)


• It
means
that
the
distribu'on
of
relevant


envionments
shared
have
the
same
mean
and
 standard
devia'n
for
MZ
and
DZ
twins


– Relevant
(has
an
impact
on
phenotype
studied)
 – Shared
(note
nothing
about
unshared
environment)


• What
are
we
taling
about?


(Dressed
alike,
Share
room,
Common
Friends,
etc.)


• Cri'cs:
“Twin
studies
is
all
crap
because
of
EEA”

• I
call
for
a
theore'cal
and
empirical
assessment



(EEA
viola'on
can
be
empirically
controlled
for)


Derks
et
al
(2006)
EEA
Model



Have
to
make
one
constrait


Trivariate
Derks
EEA
Model
(for
Twin
1)


Have
to
make
one
constrait


AYtudes
Towards
Cloning


(with
Barbara
Prainsack,
Lynn
Cherkas
and
Tim
Spector)


• UK
data
from
Tim’s
lab.


– Medical
Cloning


• Save
Life

• Eliminate
disease


– Convenience
Cloning


• Replace
Lost
Child

• Influence
Height
and
IQ


– AYtude


• Never

• Oppose

• Mechanism
for
EEA
Viola'on.


– MZ
twins
are
natural
clones.

DZ
twins
are
not.


Other
Projects


• Heritability
of
Survey
Nonresponse
(unit
and
item)

• Leadership

• Methods
(had
few,
one
survived,
one
planned)


– Confidence
Intervals
 – Proxy
Repor'ng
(just
ask
one
of
them)


• Hungarian
Twin
Registry


– Volunteer
(medical
work)
 – Popula'on
Registry
(Beast)


• Want
to
get
back
into
experimental
work


– Skin
Response
 – Startle
Response
 – Electromyography
 – Eye
Tracking
 – Hoepfully
Brain
Stuff
(maybe
one
day)


If
you
want
to
do
this
research


• Hibbing’s
team
has
given
the
ANES
to
twins


– Data
to
be
released
soon


• Add‐Health,
new
wave,
new
markers

• NSF
grant
to
train
social
science
faculty

• Boulder,
CO
Workshop

• MX
Script
Library

• Mplus
Script
Library

• Behavior
Gene'cs
Associa'on

• Interna'onal
Society
for
Twin
Studies


The
Floor
Is
Open
 for
Discussion



S'll
in
town
 (Moscow)
for
 another
week
 
If
you
want
to
 talk
shop,
or
get
 a
beverage…
 …or
just
contact
me
later:
 li#vayl@ceu‐budapest.edu