Scribd
Upload
153 views46 pages

Structural Equation Model

Structural equation modeling (SEM) is a statistical technique used to analyze relationships between observed and latent variables. It combines aspects of factor analysis and multiple regression. The document outlines SEM, discussing its two main components: confirmatory factor analysis, which deals with observed and latent variables, and path analysis, which examines relationships between observed variables. It provides an example SEM analyzing the effects of student cooperation and teacher-student interaction on student well-being. The document explains how to specify, assess the fit of, and modify SEM models using statistical software.

Uploaded by

Sarah Gracyntia
Copyright
© © All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as PDF, TXT or read online on Scribd
153 views46 pages

Structural Equation Model

Structural equation modeling (SEM) is a statistical technique used to analyze relationships between observed and latent variables. It combines aspects of factor analysis and multiple regression. The document outlines SEM, discussing its two main components: confirmatory factor analysis, which deals with observed and latent variables, and path analysis, which examines relationships between observed variables. It provides an example SEM analyzing the effects of student cooperation and teacher-student interaction on student well-being. The document explains how to specify, assess the fit of, and modify SEM models using statistical software.

Uploaded by

Sarah Gracyntia
Copyright
© © All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as PDF, TXT or read online on Scribd
You are on page 1/ 46

Structural Equation Modeling (SEM)

Mata Kuliah : Statistik Lanjut (PSPP801305 )

Whisnu Yudiana M.Psi., Psikolog


Fakultas Psikologi Universitas Padjadjaran

Hox, J. J., & Bechger, T. M. (1998). An introduction to


structural equation modeling.
Outlines
What is Structural Equation Modeling?

Why we use Structural Equation Modeling?

How to do a Structural Equation Modeling?


WHAT ?
Definition
• Structural Equation Modeling, or SEM, is a very general statistical
modeling technique, which is widely used in the behavioral sciences.
It can be viewed as a combination of factor analysis and regression or
path analysis (Hox & Bechger, 1998).

Hox, J. J., & Bechger, T. M. (1998). An introduction to


structural equation modeling.
SEM
• Two main components of SEM are presented in Figure 1.
• CFA operates with observed and latent variables, path analysis operates only with observed
variables.
Figure 1. Components of Structural Equation Modeling

(Nokelainen, 1999)
Path Analysis
Simple Linear Regression
Y = bx + a + e

b
X Y
Multiple Linear Regression

b1
X Y
b2
X
Path Analysis
• Examines how n independent (x, IV, Xi, x) variables are statistically related to a
dependent (y, DV, Eta, h) variable.
• Applies the techniques of regression analysis, aiming at more detailed
resolution of the phenomena under investigation.
• Allows
• Causal interpretation of statistical dependencies
• Examination of how data fits to a theoretical model
Path Analysis

• As stated earlier, path analysis assumes that the model


is recursive.
• Nature of causal dependency is unidirectional, like a ’one way
road’ (arc with one head ).
• If there is no a priori information available about the direction
r of causal dependency, it is assumed to be correlational (arc
with two heads ).
Path Analysis
Self Esteem

Stress Life Satisfaction

Optimism
Path Analysis
• Direct and indirect effect

Self Esteem

Stress Life Satisfaction

Optimism
Path Analysis
x IV Xi x EXOGENIOUS
Self Esteem
y DV Eta h ENDOGENIOUS

Stress Life Satisfaction

Optimism THERE regression


analyses:
1) SE + ST + OP -> LS
2) ST –> SE
3) ST -> OP
Confirmatory Factor Analysis
SEM
• Two main components of SEM are presented in Figure 1.
• CFA operates with observed and latent variables, path analysis operates only with observed
variables.
Figure 1. Components of Structural Equation Modeling

(Nokelainen, 1999.)
Confirmatory Factor Analysis

http://www.iqscorner.com/2011/04/ap101-brief-8-guest-brief-independent.html
SEM
SEM

Nachtigall et al.: (Why) Should we use SEM?


An Example of SEM: Commitment to
Work and Organization
Figure 15. Hypothesized Structural Model
CO Commitment to work and organization
PAR Participative Leadership
ELA Elaborative Leadership
ENC Encouraging Leadership
COL Collaborative Activities
CON Teacher - Student Connections
SPI Group Spirit
INC Inciting Values
VAL Job Value
INF Influence on Job
An Example of SEM: Commitment to
Work and Organization
Figure 16. Hypothesized Measurement Model
“The techniques of Structural Equation Modeling represent the future of data analysis.”
“Nobody really understands SEM.”

Why ?
Why
• To compare the model to empirical data
• to examine the fit-statistics assessing the matching of model and
data. If the model is fit, it shows the data support the theory
• The possibility of modeling complex dependencies and latent
variables
• SEM enables the analysis of latent variables and their relationships,
offering the opportunity to analyze the dependencies of
psychological constructs without measurement errors.
“The techniques of Structural Equation Modeling represent the future of data analysis.”
“Nobody really understands SEM.”

How ?
Operation Formula
Langkah SEM pada Lavaan via JASP
• Sama dengan CFA
1. Specifying Model
2. Model Summary (Factor Loading, Residual, Model Fit)
3. Modification (If Needed)
4. Re-spesification Model
5. Model fit
• Perbedaan pada:
• Memiliki Latent lebih dari pada 1
• Terdapat regression antara latent
Contoh
Data penelitian
• Outcome : Subjective Well Being In School : 8 Items (The Brief
Adolescent’ Subjective Well-Being in School Scale (BASWBSS),
developed by Tian, Wang and Huebner (2014),)
• Predictor : nventory of Student’s Perceived Learning Environment
(ISPLE) Chen chen et al, 2017)
• Student-student cooperation : 4 Items
• Teacher – Student Interaction : 4 Items
Pertanyaan Penelitian
• Apakah “Student-student cooperation” memiliki efek yang positif
terhadap “Subjective Well-Being in School“?
• Apakah ”Teacher – Student Interaction” memiliki efek yang positif
terhadap “Subjective Well-Being in School“?
• Apakah teradapat kaitan yang positif antara “Student-student
cooperation” dan ”Teacher – Student Interaction”
MENU JASP

Menguji signifikansi
dari setiap variabel/
Menetapkan teori/ model Menguji model fit
item dalam mengukur
pengukuran yang hendak diuji dari teori tersebut
konstruk/ faktor yang
hendak diukur
SEM
Student-student cooperation
• Syntax ?
Student-student cooperation
• Syntax
LSSC =~ SSC1 +SSC2 + SSC3 + SSC4
School Well-Being
• Syntax ?
School Well-Being
• Syntax
model1 <- '
SS =~ SS1 + SS2 + SS3 + SS4 +
SS6
AF =~ AF1 + AF2
SWB =~ 1*SS + AF
'
SEM
Results
Model Fit
GoF

Goodness of Fit Statistics Recommended Value


Chi-Square (df & p-value) p-value ≥ .05
Goodness of fit index (GFI) ≥ .90
Adjusted goodness of fit index (AGFI) ≥ .90
Normalized fit index (NFI) ≥ .90
Comparative fit index (CFI) ≥ .90
Root mean square residual (RMR) ≤ .08
Root mean square error of approximation (RMSEA) (p-value) ≤ .07
Hair, et.al (2010)

Sumber : R Workshop, Adiyo dkk, 2020


Lamda Parameter Estimates
Parameter Estimates
CI CI
label est se z p std (lv) std (all)
(lower) (upper)
LSSC =~ SSC1 1.000 0.000 1.000 1.000 0.911 0.567
LSSC =~ SSC2 1.057 0.082 12.927 < .001 0.897 1.217 0.963 0.578
LSSC =~ SSC3 1.071 0.078 13.669 < .001 0.917 1.224 0.976 0.630
LSSC =~ SSC4 1.083 0.078 13.862 < .001 0.930 1.236 0.987 0.645
LTSI =~ TSI1 1.000 0.000 1.000 1.000 1.022 0.704
LTSI =~ TSI2 0.758 0.055 13.708 < .001 0.650 0.866 0.774 0.520
LTSI =~ TSI3 0.933 0.058 15.963 < .001 0.819 1.048 0.953 0.615
LTSI =~ TSI4 0.942 0.060 15.621 < .001 0.824 1.061 0.963 0.600
Gama Parameter Estimate
Parameter Estimates
CI CI
label est se z p std (lv) std (all)
(lower) (upper)

SWB ~ LSSC 0.048 0.122 0.395 0.693 -0.191 0.287 0.079 0.079

SWB ~ LTSI 0.370 0.113 3.265 0.001 0.148 0.591 0.676 0.676
Theta Delta Parameter Estimate or Error
Variance
Parameter Estimates
CI CI
label est se z p std (lv) std (all)
(lower) (upper)

SSC1 ~~ SSC1 1.756 0.094 18.743 < .001 1.573 1.940 1.756 0.679

SSC2 ~~ SSC2 1.849 0.100 18.577 < .001 1.654 2.044 1.849 0.666

SSC3 ~~ SSC3 1.444 0.082 17.664 < .001 1.284 1.604 1.444 0.603

SSC4 ~~ SSC4 1.367 0.079 17.347 < .001 1.212 1.521 1.367 0.584
Rules
• Size Of Path Estimates And Statistical Significance
• provided rules of thumb suggesting that loadings should be at least .5 and
ideally .7 or higher. Loadings of this size or larger confirm that the indicators
are strongly related to their associated constructs and are one indication of
construct validity.
• Researchers should also assess the statistical significance of each estimated
(free) coefficient. Nonsignificant estimates suggest that an item should be
dropped.
• Low loadings suggest that a variable is a candidate for deletion from the
model.

(Hair, dkk, 2014)


Rules
• MODIFICATION INDICES
• A modification index is calculated for every possible relationship that is not
estimated in a model
• The modification index value would show how much the overall model χ2
value would be reduced
• Modification indices of approximately 4.0 or greater suggest that the fit could
be improved significantly by freeing the corresponding path to be estimated.
• Making model changes based solely on modification indices is not
recommended.
R Squre
Terima Kasih

You might also like