# The Dance between Accuracy and Bias - of David A. Kenny

Document Sample

```					Social Relations Model:
Multiple Variables
David A. Kenny
Types of Variables in
SRM Studies
• Personality variable
–Self variable
• Group variable
Variables
Bivariate Correlations
4 at the individual level
Individual-level Correlations
Actor-Actor
Actor-Partner
Partner-Actor
Partner-Partner
Actor-Actor
Correlation
If initially a person sees others as
Extroverted, does that person
still see others as extroverted
after interacting with them?
Not really: r = .21
Actor-Partner
Correlation
If initially a person sees others as
Extroverted, is that person seen
as extroverted after interacting
with him or her?
Maybe: r = .46
Partner-Actor
Correlation
If a person is initially seen by
others as Extroverted, does that
person see others as
extroverted after interacting
with them?
Not really: r = -.02
Partner-Partner
Correlation
If initially a person is seen by
others as Extroverted, is that
seen as Extroverted after
interacting with him or her?
Yes: r = .89
Relationship
Intrapersonal
If one person, A, initially
thinks another person, B, is
particularly extroverted, does A
still think that B is particularly
extroverted after interacting
with him or her?
Nor really: r = .23).
Relationship
Interpersonal
If one person, A, initially
thinks another person, B, is
particularly extroverted, does B
think that A is particularly
extroverted after interacting
with him or her?
Not really: r = -.15
Creating a Construct
Why?
to separate error from
relationship variance
Multiple Measures
Same measure at different times.
Different measures at the same
time.
How?
Sum or average the scores.
Create a construct or a latent
variable.
Stable versus
Unstable Variance
stable variance: variance that
correlates across different
measures of the construct
unstable variance: variance that
is unique to the specific
measure of the construct
Measurement Model
measures: All measures need to
have the same units.
Equal unstable variance in each
measure
Construct Variances
Stable Actor
Unstable Actor
Stable Partner
Unstable Partner
Stable Relationship
Unstable Relationship
Error Variance
Very often Unstable Actor and
Partner variances are very small.
There is only Unstable Relationship
variance.
of Unstable Actor, Partner, and
Relationship variances.
Example
Liking at Two Times (Curry)
Stable Unstable
Actor            .160     .029
Partner          .259     .016
Relationship     .422     .114
Error                     .159
Correlated Error
Some times, pairs of indicators share
method variance.
Same time
Same instrument
Need to remove correlated error
effect in computing correlations
between two constructs.
A Personality
Variable with
Variable
Extroversion (personality
variable) with Smiling
Actor Personality Variable
Correlation: If Dave is extroverted,
does Dave smile more?
Partner Personality Variable
Correlation: If Dave is extroverted,
do others smile more at Dave?
A Personality Variable at
the Relationship Level
Compute the product of actor’s
personality X partner’s
personality (both centered) or
alternatively the absolute
difference.
Correlate with relationship effect.
Self Variable as a
Special Personality
Variable
measurement in which actor
and partner are the same
person.
Can correlated with actor and
partner effects.
Group Variable
Same score for all group
members.
Examples
gender
experimental condition
Tests
level
variances