Scribd
Upload
59 views15 pages

January 19 Introduction To SEM

This document provides an introduction to structural equation modeling (SEM) and discusses several key concepts: 1) It describes different SEM models including confirmatory factor analysis, path models with latent variables, and models of developmental trajectories. 2) It discusses how to specify SEM models based on research questions and hypotheses, and how to represent direct and indirect effects between variables. 3) It covers issues related to model identification, estimation, fit, assumptions, missing data handling, testing moderation, and comparing models across groups.

Uploaded by

tua87538
Copyright
© Attribution Non-Commercial (BY-NC)
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as PPT, PDF, TXT or read online on Scribd
59 views15 pages

January 19 Introduction To SEM

This document provides an introduction to structural equation modeling (SEM) and discusses several key concepts: 1) It describes different SEM models including confirmatory factor analysis, path models with latent variables, and models of developmental trajectories. 2) It discusses how to specify SEM models based on research questions and hypotheses, and how to represent direct and indirect effects between variables. 3) It covers issues related to model identification, estimation, fit, assumptions, missing data handling, testing moderation, and comparing models across groups.

Uploaded by

tua87538
Copyright
© Attribution Non-Commercial (BY-NC)
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as PPT, PDF, TXT or read online on Scribd
You are on page 1/ 15

Introduction to

Structural Equation Modeling


Research Questions & SEM
• Confirmatory Factor Analysis.

• Path Models with Latent Variables.

• Models of Developmental Trajectories.

• Model Differences between Groups.


r1
CFA Nervous
Not Calm r2

Tired Out r3

Anxiety Effort r4

Can’t Sit Still r5

Restless r6

Worthless r7
Depression
Can’t Cheer Up r8

Depressed
r9
Hopeless
r10
Path Model with Latent Variables

Family Conflict

Child
Observer Child
Parent
Depression

Child Age
Child Observer
Child Gender
Parent
Latent Growth Curve Model

Family Conflict
Wave 1
Trajectory
Mother Child ASB
Child
Father

ASB ASB
Child Self Esteem Wave 2 Wave 4
Wave 1
ASB
Wave 3
Mother Child
Father
Think in Terms of Models
• Strictly Confirmatory.
• Alternative Models.
• Model Generation.
Testing Alternative Models

Romney, Jenkins, & Bynner (1992) Human Relations 45, 165-176


Model Specification

• The researcher’s hypotheses are expressed in the


form of a SEM.

• What are the variables that effect a particular


phenomenon of interest?

• What are the pathways of those effects?


(direct and indirect)
Equivalent Path Models
Model Identification
• Is it theoretically possible to calculate a unique estimate
for all of the model’s parameters?

• Two key issues


– The number of model parameters can not exceed the number of
“observations” – number of degrees of freedom
• Over-identified, Just-identified, Under-identified
– Every unobserved variable must be assigned a scale.

• “Observations” are the variances and covariances of the


measured variables (correlation matrix).
– v(v + 1)/2, where v is the number of observed variables
– Have 15 “observations” in Romney et al.
Model Estimation
• Use a model-fitting program to derive estimates of model
parameters (Amos, EQS, LISREL, Mplus).

• Maximum Likelihood (ML) is by far the most widely used


estimation procedure.

• However, ML assumes multivariate normality. If the data


are not MVN, then other procedures can be used.
Assessing Model Fit

• Determine how well the model accounts for the


observed variances and covariances of the measured
variables.

• Fit Indices: 2, GFI, CFI, TLI, RMSEA, SRMR, and many
others.

• These indicate only the overall or average fit of the


model, and do not indicate whether the results are
theoretically meaningful.
Assumptions within SEM
• Large samples: try to obtain a 10:1 ratio of number of
subjects to number of model parameters.

• Variables are typically at interval or ratio level of


measurement, although not necessarily.
– Mplus program designed to analyze categorical variables

• Approximately multivariate normal distribution.

• Missing data are MAR.


Dealing with Missing Data in SEM
• Listwise deletion: assuming MCAR

• EM Algorithm and Multiple Imputation methods are available


in most SEM software programs

• Common approach is to use Full Information Maximum


Likelihood (FIML) estimation
• Uses all of the raw data, regardless of the amount of missingness for
any given case.
• Partitions the sample into subsets of cases having the same pattern
of missingness.
• All available statistical information is drawn from each subset; all
cases are used in the overall analysis.
Testing for Moderation in SEM
• Interactions between observed variables
– Create product terms to include in path model

• Interactions between latent variables


– Kenny-Judd method, need to use nonlinear constraints in
measurement model
– Ping method, involves calculation of loadings for product
indicators, fixed parameters.

• Test group differences in model using 2 differences when


constraining parameters to be equal across groups.

You might also like