Structural Equation Modelling (SEM) Part 2

This presentation is an introduction to the concept and theory of Structural Equation Modelling.

1. Structural Equation Modelling
(SEM)
An Introduction (Part 2)

2. SEM: Basic Concepts
•MeasuredVariableorIndicatorVariable
•LatentVariable
•MeasurementModel
•StructuralModel

3. Basic Concepts: Measured Variable/Indicator
•Measuredvariable(s)arethevariablesthatareactuallymeasuredinthe
study.
Latent Variable
Measured Variable 1 Measured Variable 2 Measured Variable 3

4. Basic Concepts: Latent Variable
•Intangibleconstructsthataremeasuredbyavarietyofindicators
(moreisbetter!)
Latent Variable
Measured Variable 1 Measured Variable 2 Measured Variable 3

5. Basic Concepts: Measurement Model
•Themeasurementmodelcanbedescribedasfollows.Itshowsthe
relationshipbetweenalatentvariableanditsmeasured
items(variables).
Latent Variable
Measured Variable 1 Measured Variable 2 Measured Variable 3

6. Basic Concepts: Structural Models
•OftenusedtospecifymodelsinSEM
Causalflowisfromlefttoright;toptobottom
•Straightarrowsrepresentdirecteffects
•Curvedarrowsrepresentbidirectional“correlational”
relationships
•Ellipsesrepresentlatentvariables
•Boxes/rectanglesrepresentobservedvariables

7. Example:Structural Models

8. Variants of Structural Equation Modelling
•ConfirmatoryFactorAnalysis(CFA)
•PathAnalysiswithobservedvariables
•Pathanalysiswithlatentvariables

9. Confirmatory Factor Analysis
“Measurement Model”
•Testsmodelthatspecifiesrelationshipsbetweenvariables(items)and
factors
Andrelationshipsamongfactors
•Confirmatory
Becausemodelisspecifiedapriori

10. Example: Oblique CFA Model

11. Confirmatory vs. Exploratory Factor
Analysis
•InCFAthemodelisspecifiedapriori
Basedontheory
•EFAisnotamemberoftheSEMfamily
Includesaclassofproceduresinvolvingcentroids,principalcomponents,and
principalaxisfactoranalysis
Doesnotrequireapriorihypothesisaboutrelationshipswithinyourmodel
Inductivevs.deductiveapproach
Morerestrictionsontherelationshipsbetweenindicatorsandlatentfactors

12. Example: Oblique EFA Model

13. Observed Variable Path Analysis (OVPA)
•Testsonlyastructuralmodel
Relationshipsamongconstructsrepresentedbydirectmeasured
(observedvariables)
i.e.,each“box”inmodelisanidem,subscale,orscale
•Analogoustoaseriesofmultipleregressions
But,withMR,wewouldneedkdifferentanalyses,wherekis#of
DVs
WithSEM,cantestentiremodelatonce

14. Example: OVPA

15. Latent Variable Path Analysis (LVPA)
•Simultaneous test of measurement and structural parameters
•CFA and OVPA at same time
•LVPA models incorporate….
•Relationships between observed and latent variables (i.e., measures and factors)
•Relationships between latent variables
•Error & disturbances/residuals

16. Example: LVPA

17. Data Considerations
SampleSize
•SEMisalarge-sampletechnique
•TherequiredSamplesizeneededdependson….
Complexityofmodel
Ratiosofsamplesizetoestimatedparametersrangingfrom
5:1to20:1(Bentler&Chou,1987;Kline,2005)
DataQuality
Largersamplesfornon-normaldata

18. Looking for Online SEM
Training?
Contact us: [email protected]
Visit: http://tinyurl.com/costarch-sem
www.costarch.com