Dyadic Data Analysis Using Multilevel Modeling with R

Dyadic Data Analysis with R

DATIC 2018: Workshop Curriculum

June 25 – June 29, 2018

Instructors: Randi L. Garcia and David A. Kenny

Should I Attend?

We presume the variables you want to analyze are from two people who are measured on the SAME variables. For instance, you have data from husbands and wives who are both measured on satisfaction, commitment, and equity. If you have data from two people and say you have social support measured from one person and pain from the other person, the methods that we describe would be of little benefit to you. Second, we presume that each person is a member of one and only one dyad. If you have data where all possible dyads are formed from a group of people or people have dyadic relationships with the same person (e.g., a teacher with students), much of what we cover would be of little benefit to you.

Overview

This one-week workshop on Dyadic Data Analysis will be held at the University of Connecticut from Monday, June 25, through Friday, June 29, 2018. The workshop focuses on the analysis of dyadic data when both members of a dyad are measured on the same variables. All analyses will use multilevel modeling in R via the RStudio graphical interface. Other methods could be used (e.g., Structural Equation Modeling) and other software can be used to do multilevel modeling (e.g., SPSS, HLM, MPLUS), but in this workshop we focus on R, mainly because it is free and therefore widely accessible. Moreover, we shall make available setups for HLM, SPSS and SAS for some of the analyses that we present. No prior familiarity with R is required, but if you have never used R and want to gain a general proficiency working with data in R, we encourage you to take the two-day DATIC Intro to R and RStudio workshop held on Thursday, June 20, through Friday, June 21, 2018 (or Thursday, June 7, through Friday, June 8, 2018).   In this Dyadic Data Analysis workshop we will focus almost solely on running dyadic models in R.

Among the topics to be covered are the vocabulary of dyadic analysis, non-independence, data structures, and the Actor-Partner Interdependence Model. We will also discuss mediation and moderation of dyadic effects and, on the fourth day, you will choose from one of two break-out sessions: 1) the analysis of over-time dyadic data (e.g., growth curve models) or 2) dyadic data analysis with SEM using the lavaan R package (e.g., Actor‑Partner Interdependence Model and Common Fate Model). Participants will learn how to analyze dyadic data and to interpret the results from their analyses.  The discussion of over‑time data is limited to one day and so the workshop should not be construed as workshop on longitudinal dyadic analysis.

It is expected that participants have a working knowledge of multiple regression. We extensively discuss the use of multilevel modeling, but participants need not be familiar in advance with this technique. Instruction consists of lecture, computer workshops, and individualized consultations. The emphasis will be practical with minimal emphasis on statistical theory, but those seeking more statistical information can arrange an individualized session with the instructors.  A teaching assistant will also be available for consultation and assistance.

The computer workshop exclusively uses R. Several R setups and web apps will be described for data organization, analysis, and write-ups. Although prior experience with the program is not required, it is desirable. Also in meetings with individual participants, alternative programs and approaches (e.g., SPSS, HLM, MLwiN, SAS, Amos, and Mplus) can be discussed. Neither instructor has much familiarity with Stata.

Participants are encouraged, but not required, to bring their own data so that they can apply these new methods to their own data. They should contact the staff beforehand to ensure that data are appropriately formatted for analysis. SPSS (i.e., .sav) or Excel (i.e., .csv or .xlsx) files are acceptable.
Logistical details: The Dyadic Analysis workshop runs Monday-Thursday from 9-5 (with individual consultations available from 3:30-5:00) and Friday from 9-12. A light continental breakfast is provided every morning. There is also a closing group lunch on Friday from 12:00-1:00 pm.

Benefits:

The following materials and events are included in the cost of the workshop:

  • Binder with workshop outline, computer setups and outputs
  • R programs to assist in the write-ups of the results of a dyadic analysis
  • Meals: Continental breakfast each morning, complimentary luncheon Friday

Recommended Advanced Reading:

R for Data Science by Grolemund and Wickman (2017), published by O’Reilly

  • Participants would benefit from reading at least Chapters 3 and 5 of the book in advance. It is freely available online: http://r4ds.had.co.nz/.

Dyadic Data Analysis by Kenny, Kashy, and Cook (2006), published by Guilford Press

  • Participants would benefit from reading Chapters 1, 2, 3, 4, 7, and 13 of the book in advance. Copies of these chapters and other readings will be distributed at the workshop.

Schedule:

Monday – Thursday

o 8:30am – 9:00am: Continental Breakfast

o 9:00am – 3:30pm: Workshop, including computer workshops

o 3:30pm – 5:00pm: Work on one’s own data;  optional homework; individual meeting time with workshop instructors

Friday

o 8:30am – 9:00am: Continental Breakfast

9:00am – 12:00pm: Workshop

o 12:00-1:00pm: Complimentary Group Luncheon

 

Day 1: Introduction to the Dyadic Data

  • Definitions: Distinguishability, Types of Variables, and Nonindependence · Estimating, Testing, and Interpreting Nonindependence
  • Management of Dyadic Data: Data Structures and Data Restructuring · Using Multilevel Modeling with Dyadic Data

Day 2: The Actor-Partner Interdependence Model

  • Estimation of Models for Indistinguishable Dyads · Estimation of Models for Distinguishable Dyads
  • Power Analysis
  • Testing Specialized APIM Models
  •  Testing for Indistinguishability
  • Actor-Partner Interaction
  • APIM apps

Day 3: The Actor-Partner Interdependence Model:

  • Complications
  • Mediation of the APIM Effects
  • Moderation of the APIM Effects

Day 4: Over-Time Dyadic Data

Break-out Session 1:

  • Repeated Measures Analysis
  • Data Structure and Lagged Values
  • Growth-curve Models
  • Stability-Influence Model

Break-out Session 2:

  • Using SEM with Dyadic Data: Introduction
  • Actor-Partner Interdependence Model
  • Common Fate Model
  • Mediation Models

Day 5: Additional Topics (Instruction Ends at Noon- Group Lunch from 12:00-1:00)

  •  Presenting Dyadic Data Analyses
  • Using GEE and Generalized Linear Mixed Models with Dyadic Data
  •  Alternative Dyadic Designs
  •  Alternative Dyadic Models
  •  DyadR

Register for the Dyadic Analysis workshop here.