Mild introduction to Structural Equation Modeling (SEM) using R

Description:

When working with data, we often want to create models to predict future events, but we also want an even deeper understanding of how our data is causally connected or structured. In this workshop, we explored the connectedness of data using structural equation modeling (SEM) with the {lavaan} package.

The packages used in this video is lavaan, semPlot, MPsychoR and corrplot. We recommend that you have these packages installed prior to the starting if you want to follow activly. Below is a line of code you can use.

The speaker, Ronny Scherer, received his PhD from Humboldt-Universität zu Berlin in 2012, and is currently working at the Centre for Educational Measurement, University of Oslo (CEMO). His research and teaching cover structural equation modeling, meta-analysis, computer-based assessments, and multilevel analysis. Ronny is coordinating the Measurement Models course at CEMO.

Slides and code used in the video can be found here:
Slides:
Code:

Timestamps:
0:00:00 Start
0:00:39 Welcome and introduction to the workshop
0:05:57 Structural equation modeling—Why? Definition and advantages
0:14:45 Structural equation modeling—What? Examples from different disciplines
0:22:45 Structural equation modeling—How? Steps taken in SEM
0:35:54 Illustrative example—Model 1: Linear regression
0:41:48 Implementation of Model 1 in lavaan
1:09:53 Testing the equality of (unstandardized) regression parameters in Model 1
1:16:26 Illustrative example—Model 2: Mediation model
1:20:00 Implementation of Model 2 in lavaan
1:30:07 Illustrative example—Model 3: Confirmatory factor analysis
1:34:58 Implementation of Model 3 in lavaan
1:48:54 Illustrative example—Model 3b: Confirmatory factor analysis modified
1:51:28 Implementation of Model 3b in lavaan and model comparison
2:04:33 Illustrative example—Model 4: Structural equation model
2:05:58 Implementation of Model 4 in lavaan
2:19:15 Illustrative example—Model 5: Multi-group structural equation model
2:21:00 Data issues in SEM—What if’s and possible solutions

[duncanunwin]

The best introduction to SEM I''ve seen. Thanks.

[scarlettzhang]

Thank you very much to all presenters! This is a very newbie-friendly video to learn SEM using R! Everything is well organized and well explained! Looking for more good videos:)

[clementinetine]

Thank you so so so much for this AMAZING tutorial!!! This is actually my second time trying to use R, and this video was a huge help in understanding the basic logic behind conducting SEM using R. I could even code the mediation model by myself in 3 minutes, which made me very happy :D You are a great educator

[amiraalaamri]

Thanks a lot for this great presentation. It is very informative.

[jayh]

Wonderful presentation! Thank you so much! :)

[JeffNijsse]

Thanks Simon Pegg! Its been a pleasure learning SEM from you. Very well thought out and delivered workshop 👍

[tomatolz]

It's a great great course. Thank you!

[theresae]

Hi, thanks for the great tutorial. I got stuck with all the other explanations, but now I had the chance to get one step further. Great Job!!

[xuyang]

Hello, Author. Could you tell me how to get the residual vairances of a MSE by lavaan()? Thanks

[aysusecmen]

Thank you so much, great workshop. In your last slide, you are talking about multilevel SEM. Is there an easy syntax example you can share? Say I have longitudinal data, we have the same measurements of openness, agreeableness and prejudice, but this time the data is so that the measurements have been collected across multiple participants through time. For simplicity we can assume we have t1 ...tn, n time stamps for all measurement for each participant. How the model 4 syntax can be modified to incorporate this? (we would be using linear mixed effect models with random intercepts (1|participant_id)