Structural Equation Modeling

29 April - 3 May 2019

Magdalena Steiner, University of Fribourg
Prof. Sven Bacher, University of Fribourg

Dr James Grace, US Geological Service, Lafayette, Luisiana

Natural systems are dynamic and involve many factors which interact and influence each other. This influence can be direct or indirect where effects can be «hidden» and the investigator will miss the importance of such background factors.

Structural Equation Modeling (SEM) is a statistical method to evaluate a priori theory-based models and a statistically more comprehensible alternative to null hypothesis testing. While other statistical techniques such as CCA, PCA or NMDS are rather used to explore multivariate datasets, SEM is used in confirmatory analyses and the evaluation of multivariate hypotheses. It has the capability to look at and distinguish between effects of different factors in one system which makes it possible to investigate multiple hypotheses at the same time. Therefore SEM is a powerful complementary skill to other multivariate techniques for researchers in ecology.

In this workshop the participants will be introduced to the topic and different methods of SEM. The course is a primer on structural equation modelling (SEM) and confirmatory path analysis, with an emphasis on practical skills and application. Theoretical content will be visualized and strengthened by practical exercises. Participants have the possibility to practice on prior to the course prepared datasets but are also encouraged to bring their own data. The used course software is R.

A basic statistics knowledge as well as basic skills in R are a precondition for this workshop

Pre-class Questionnaire on Background and Interests

Pre-class Practice Example using lavaan package in R

Day 1
Overview

• What is Structural Equation Modeling?
• The Three Most Common Implementations in R

Getting Started

• An Introduction to “lavaan”
• Example 1: Observed-Variable Model Implemented in lavaan

SEM Essentials

• Model Anatomy and Terminology
• Estimation
• Path Rules
• Model Evaluation

In-class Practice Exercise on Model Evaluation

Background Reading:

• Grace 2006 Structural Equation Modeling and Natural Systems – Chapter 2
• lavaan Syntax Guide

Day 2

Illustrations of Distinctive Features of SEM

• The Test of Mediation
• SEM versus Multiple Regression
• SEM versus ANOVA and ANCOVA
• In-class Practice Exercise Involving Mediation Hypothesis

Modeling More Abstract Hypotheses

• Modeling with Latent Variables
• In-class Practice Exercise with Latent Variables
• Modeling with Composite Variables

Background Reading:

• Prugh et al. 2018. Nature Climate Change, Aug 20:1.
• Daou and Shipley. 2019. Ecology 100: e02549

Day 3

In-class Practice Exercise with Composites

Additional Topics

• Interpreting Coefficients
• Working with Categorical Predictors
• Additional lavaan Options
• Complex Survey Data and Multi-level Modeling
• Reciprocal Effects/Causal Loops

Longitudinal Data

• Time-Step Models
• Time-Trajectory Models

Background Reading:

• Gross 2016 Nature Highlight
• Grace et al. 2016 Nature

Day 4

Introduction to Local Estimation and Evaluation

• What is Local Estimation and Evaluation (LEE) and Why Use It?
• Evaluating Model Structure in LEE

Introduction to the "piecewiseSEM" (PWSEM) Package

Introduction to 'Artisan' Implementations

Incorporating Non-Gaussian Responses

• Using piecewiseSEM
• Using an Artisan Approach
• In-class Practice Exercise using piecewiseSEM

Accommodating Random Effects

• Using piecewiseSEM
• Using an Artisan Approach

Non-linear Linkages

Background Reading:

• Lefcheck 2016 Methods in Ecol. & Evol. 7:573–579.

Day 5

Wrap up of Formal Lectures

Open Lab Time

Presentations and Discussions

University of Fribourg

2.5

2.5 ECTS with full attendance and active participation

When?

29 April - 3 May 2019

9.00 - 17.00

Where?

University of Fribourg (CH)

Please note the different rooms:

Monday 29 April: Conference room 0.36, PER17 building - Chemin du Musée

Tuesday 30 April: Seminar room 0.367, PER05 building - Chemin du Musée

Wednesday 1 May: Conference room 0.36, PER17 building - Chemin du Musée

Thursday 2 May: Conference room 0.36, PER17 building - Chemin du Musée

Friday 3 May: Seminar room 0.109, PER04 building - Rue Albert-Gockel, 3

Questions?

Marta Bellone

@: ecologie-evolution(at)cuso(dot)ch

Phone: +41 (0)26 300 88 91

Reimbursement:

PhD students of the DPEE are eligible for reimbursement of incurred travel expenses by train (half-fare card, and 2nd class). Please send the original tickets along with the reimbursement form to:

 Marta Bellone  
 Doctoral Program in Ecology and Evolution
 PER04 building
 University of Fribourg
 Chemin du Musée 10
 CH-1700 Fribourg

NO reimbursement of your individual meal expenses

Regarding reimbursement of accomodation, please contact the coordinator of the doctoral program (ecologie-evolution(at)cuso.ch) BEFORE the beginning of the course. NO reimbursement of accomodation without the agreement in advance of the course of the coordinator of the doctoral program.

CUSO PhD students: through your MyCUSO account. 

External participants (non-CUSO PhD students, post-docs, etc...): use the icon "registration" at top of page and the last gray box "non-CUSO student" ("personne hors myCUSO").

Deadline for registration: April 7, 2019. Priority will be given to CUSO DPEE PhD students until March 24, 2019.

Registration fees:

CUSO DPEE PhD Students: free

Other CUSO members: 165 CHF (it includes all the 5 collective lunches, coffee breaks and a social dinner)

External participants (i.e. from a non-CUSO university): Please contact ecologie-evolution(at)cuso(dot)ch before registering!

16