Stuff I Have Done and Am Doing Now David A. Kenny.

Stuff I Have Done and Am Doing Now

David A. Kenny

Stuff I Have Done

Social Relations ModelActor-Partner Interdependence Model

2

Social Relations Model

groups or teamsdirected dyadic data: A’s perception of B & B’s of A

gathered from all or mostly all pairs 3

Round Robin Design

1 2 3 4 5 61 - x x x x x

2 x - x x x x

3 x x - x x x

4 x x x - x x

5 x x x x - x

6 x x x x x -

4

SRM Equation

For actor i with partner j in group k (e.g., how intelligent i sees fellow member j in group k):

Xijk = mk + aik + bjk + gijk

5

actor

group partner

relationship

Reciprocity Equations

Xijk = mk + aik + bjk + gijk

Xjik = mk + ajk + bik + gjik

6

A Very Messy Multilevel

ModelThree level model: group, individual, and observation

Two crossed random variables at the individual level (actor and partner)

Linkage across them: Actor and partners the same people

7

EstimationANOVA modelsSnijders approach creation of many dummy

variables many constraints on the tau

matrix8

SRM Example: Leadership

Group

Actor Part.

Relat. Error

Leadership

.00 .09 .43 .19 .29

Variance Partitioning (proportions)

Actor-Partner(Generalized)

Relationship

(Dyad)

Leadership .14 .03

Reciprocity (correlations)

9

Random and Fixed Effects

Most SRM analyses (like the above) focus on the random effects.

Estimation of fixed effects at the group and individual level within the SRM is relatively straight-forward.

Estimation of fixed effects at the relationship level is not so simple.

10

Research Question

Metaperception of Liking: How much person 1 thinks 2 likes 1 or P12

Two dyadic predictors:How much person 2 likes person 1 or A21 (accuracy)

How much 1 likes 2 or A12 (assumed

reciprocity) 11

A’s Perception of B’s Liking of A

Actor Partner Interdependence Model

Bias

Accuracy

A’s Liking of B

B’s Liking of A

12

A’s Perception of B’s Liking of A

Actor Partner Interdependence Model

B’s Perception of A’s Liking of B

Bias

Bias

AccuracyAccuracy

A’s Liking of B

B’s Liking of A

13

APIM in Groups: GAPIM (Kenny &

Garcia)Terms

Actor: Effect of own XPartner: Effect of others’ XActor similarity: Similar of the actor to others.Others’ similarity: How are the others

Terms can be combined to createDiversityGroup compositionFrog pond effects 14

Fixed Relationship

EffectsCould enter in dyadic variables A12 and A21 as predictors within a multilevel model.

Two problemsEstimation messiness of the Snijders approachSome of the effects of dyadic predictors will be at the individual and group levels.To obtain a “pure” dyadic measures is very messy.

15

StrategyRemove from the data all of the

group and individual variance.What remains is purely dyadic, i.e.,

relational.

Akin to the old-fashioned “within” approach for the hierarchically nested design.

16

Curry & Emerson 6 groups of 8 personsLiking and metaperception of liking

measured at five timesData from Week 1, the first

measurement, are used.

17

EstimatesTerm b (SE) Confidence IntervalAR .363 (.036) .293 to .432Acc .084 (.036) .014 to .153

http://davidakenny.net/doc/sre.pdf

18

Current Work:DataToText

Methodologists need to become more consumer oriented.

Computer output needs to be presented in ways that are more user friendly.

What about abuse?19

Current Work:DyadR

Developed a series of programs for dyadic analysis in R.

Use Rstudio’s shiny interface.Programs do not require that the

user instal R.Provide the usual computer

output, text, tables, and figures.20

DyadR Programs

Restructuring Dyadic Datadavidakenny.net/RDDD.htm

Analysisdavidakenny.net/DyadR/DyadRweb.htm

Tests of distinguishabilityAPIM power analysesAPIM analysesOther stuff 21

Woman’s Perception of Man’s

Rel. with Child

DyadR APIM Example

Man’s Perception of Woman’s Rel.

with Child

Female Bias

Male Bias

Male Accuracy

Female

Accuracy

Data gathered by Linda Acitelli and the model adapted from West & Kenny Truth and Bias model.

Woman’s Rel. with Child

Man’s Rel. with Child

22

Thank You!