Structural Equation Modeling: SEM

. .

2556

SEM

SEM
(Structural
Equation Modeling: SEM)
(testing)
(estimate)
(causal relationships)


(confirmation)
(exploration)
(theory
testing) (theory
building)
Bollen, K A., & Long, S. J. (1993). Testing Structural
Equation Models. SAGE Focus Edition, vol. 154.
Retrieved March 18, 2010, from
http://en.wikipedia.org/wiki/Structural_equation_mod
eling#cite_note-3

Theory Testing VS Theory Building

(theory testing)
(deductive)
(quantitative research)

(causal model)
(fit)

(confirmatory factor analysis)
(theory building)
(inductive)
(qualitative research)
(free parameters)

(exploratory factor
analysis)

SEM
What is Structural Equation
Modeling? (http://www2.gsu.edu/~mkteer/sem.html)
SEMis a largely confirmatory, rather than

exploratory, technique. That is, a researcher are more

likely to use SEM to determine whether a certain
model is valid., rather than using SEM to "find" a
suitable model--although SEM analyses often involve a
certain exploratory element.

Garson (2009)

SEM is usually viewed as a confirmatory

rather than exploratory procedure

(theory testing)

Garson, G.D. (2009). Structural

Equation Modeling. Retrieved March
18, 2010, from http://faculty.chass.
ncsu.edu/garson/PA765/structur.htm

SEM
2
(exogenous variables) (endogenous variables)

/
/
(independent variable) (dependent variable)

/ (regression)
/

(regression)

SEM

/ (structural model/structural equation
model) (causal relationship)
( )
(recursive and linear additive) (nonrecursive and linear additive) (measurement model)
()

(factor analysis model)
(measurement model) (path
diagram) (structural model)

SEM --- Path analysis + Factor analysis

(path analysis)
(factor analysis)

(path analysis)

(factor
analysis)

SEM


1)
2)
3)
4)

5)
6)

- (chi-square) p-value 0.05

- /df 2.00
goodness of fit index: GFI, adjusted goodness of fit
index:AGFI, comparative fit index: CFI
0.90 1.00
standardized root mean squared residual: standardized
RMR, root mean square of error approximation: RMSEA
0.05
critical n: CN 200
largest standardized residual -2 2

( )

(path analysis)

!!

(2548). .
, 1(2), 9-18.

(path analysis)
1)
2)
3)
4)
5)

(interval)
2
(identified model)
(over just identified model)

(2548). .
, 1(2), 9-18.

(SES)
(IQ)
(nAch)




(GPA)

1)
2)
3)

4)

1)

(direct effects)
(indirect effects)




(hierarchical linear model: HLM)




(2548). .
, 1(2), 9-18.

(observed variable) (latent or
unobserved variable)

STRUC

THINK

VISION

ACTIV

KNOW

CHILD

RESOU
ICT

LEARN
PAR

PROCESS
POTENT

NT
CHAR

EFFECTIV
DEFF
MORN
DEVEL
CHANG

SATIS

ACADEM
ENVIR
EXPEC
POSIT

LO
MOTIV

TRUST

SHARE

. (2551). : .

.

 (measurement model) (structural

equation model)
( )
(exogenous latent variable) 1) POTENT (
) 5 (1) (STRUC) (2)
(VISION) (3) (KNOW) (4) (RESOU)
(5) (ICT) 2) ACADEM (
) 4 (1) (DEFF) (2)
(MORN) (3)
(DEVEL) (4) (CHANGE)
(endogenous latent variable) 1) PROCESS (
) 5 (1) (CHILD) (2)
(ACTIV) (3) (THINK) (4)
(LEARN) (5) (PAR) 2) ENVIR (
) 5 (1) (EXPEC) (2) (POSIT) (3)
(TRUST) (4) (SHARE) (5)
(MOTIV) 3) EFFECTIV ( ) 4
(1) (NT) (2) (CHAR) (3) (SATIS)
(4) (LO)



POTENT
ACADEM
PROCESS
ENVIR
EFFECTIV
PROCESS
ENVIR
(intervening or mediating variable
or mediator) ACADEM
EFFECTIV
PROCESS

PROCESS
POTENT
EFFECTIV
ACADEM
ENVIR

STRUC

THINK

VISION

ACTIV

KNOW

CHILD

RESOU
POTENT

DEFF

ACADEM

DEVEL
CHANG

PAR

NT

PROCESS

ICT

MORN

LEARN

EFFECTIV

CHAR
SATIS

ENVIR
EXPEC
POSIT

LO
MOTIV

TRUST

SHARE

(simultaneous equation model) (model
goodness of fit test) (Structural Equation
Model: SEM)
SEM LISREL
Mplus SPSS


(Structural Equation Model: SEM)
SEM Karl G. Joreskog .. 1960




Kuhnel (2001 , 2548)
........

(2548). .
, 1(2), 9-18.

MRA, ANCOVA
ANOVA

t-test
Hotelling T2

MANOVA

MMRA, MANCOVA
PATH ANALYSIS

FACTOR, CLUSTER
ANALYSIS

NONPARAMETRIC
STATISTICS

LISREL or SEM

HLM

(2548)
3
t-test ztest

(analysis of variance: ANOVA) t-test

ANOVA
(multiple regression analysis:
MRA) ANOVA ANCOVA (analysis of covariance) MRA
(multivariate statistical analysis)
ANOVA, ANCOVA, MRA (multivariate
analysis of variance: MANOVA)
(multivariate
multiple regression analysis: MMRA)
(path analysis)


(hierarchical linear model : HLM)

(dependence
statistical technique)
(2548). .
, 1(2), 9-18.


NONPARAMETRIC
STATISTICS

LISREL or SEM
MRA, ANCOVA
ANOVA

t-test
Hotelling T2

MANOVA

MMRA, MANCOVA
PATH ANALYSIS

HLM

FACTOR, CLUSTER
ANALYSIS

(non parametric statistics)






(interdependent statistical technique)

(factor analysis)
(multidimensional scaling)


SEM
(multi-level causal
model)
(longitudinal factor analysis model)
(multiple population model)
(latent growth curve
model) (Joreskog and
Sorbom , 1996 , 2548)

SEM
(Joreskog,
et al, 1999 , 2548)
SEM

(2548). .
, 1(2), 9-18.


(2548) SEM


t-test


SEM SEM






SEM ANOVA,
MANOVA
(2548). .
, 1(2), 9-18.


SEM



SEM

1.

SEM




20
(2548). .
, 1(2), 9-18.


2.

- -
H0 : =
chi-square test

Meehl and Waller (2002 , 2548)

(SEM)

Joreskog and Sorbom (1996 , 2548)
(analysis
sample) (validated sample)

(2548). .
, 1(2), 9-18.


3.




SPSS
SEM





(2548). .
, 1(2), 9-18.


SEM
SEM

SEM SEM
SEM

SEM

SEM
SEM


SEM
SEM

(2548). .
, 1(2), 9-18.

+ ...

2


--
3
20
(
)

SEM LISREL, MPLUS, AMOS, EQS

, /df, GFI, AGFI, CFI, standardized RMR, RMSEA, CN
largest standardized residual, Q-plot

---

 SEM ....
....
2
review +
....


SEM


review
2

..... 2
......

.....

....

....

1, 2,3, 4
..................................

...
Warning: Indicator variables cannot be combined
arbitrarily to form latent variables. For instance,
combining gender, race, or other demographic
variables to form a latent variable called "background
factors" would be improper because it would not
represent any single underlying continuum of meaning.
The confirmatory factor analysis step in SEM is a test of
the meaningfulness of latent variables and their
indicators but the researcher may wish to apply
traditional tests (ex., Cronbach's alpha) or conduct
traditional factor analysis (ex., principal axis factoring)
as well. Rex B. Kline

3.
4.
5.

6.
7.

11.
12.
13.
14.

1
1
1
1

17.
8

17

16.

33

15.

33

9.

33

10.

taxonomy revised 2001

(2553)

(2553)

(2553)
(2553)

(2553)

(2552)

(2552)

(2544)

Rock (2009)

Leonard (2008)

Thompson (2005)

(2547)
(2552)

Blooms

8.

Kahan(2002)

Kapur(2007)

Thomus (2005)

(2005)

Leimbach (2006)

Zaccaro& Bennis (2004)

Williams (2005)

ManningandRobertson (2002)

Thompson(2000)

Marinelli (1998)

Piciacchia(1996)

Nanus (1992)

Fisher and Wilmore (1993)

Braun (1991)

Lesoure (1990)

Westley & Mintbarg (1989)

Kantabutra and Avery (2004)

2.

Sashkin& Sashkin (2003)

1.

Hickman& Silva (1984)

Bennis and Nanus (1985)

.. review

14
5

(Synthesis)

 ...

---

 ... SEM
http://ednet.kku.ac.th/~edad/OLD/Dissertation_Pdf/Samrit.pdf
http://ednet.kku.ac.th/~edad/OLD/phenporn.pdf
http://ednet.kku.ac.th/~edad/OLD/Jirawan_PhD[1].pdf
http://ednet.kku.ac.th/~edad/OLD/Nikanchala_thesis.pdf
http://ednet.kku.ac.th/~edad/OLD/Nikanchala_thesis.pdf
http://www.mbuisc.ac.th/phd/thesis/Supphakant.pdf
http://www.mbuisc.ac.th/phd/thesis/wongwarin.pdf

 SEM.

SEM




..................

Some Questions
How many observed variables are recommended?
Typically, each variable in the model is conceptualized as a latent one,
measured by multiple indicators. Multiple indicators are developed for each
latent variable, with at least two and preferably three or more indicators per
latent variable.
Indicator variables are observed variables, sometimes called manifest
variables or reference variables. Items in a survey instrument may be
indicators, for instance. Four or more indicators are recommended, though
three is acceptable and common practice. As few as two indicators or even
a single indicator may be acceptable if the researcher is confident in the
measure's validity and reliability.
What is the problem with having less than three indicators per variable?
How many are recommended? (Point 3) Four are recommended, three is
okay, two is problematic, with one measurement error cannot be modeled.
Models using only two indicators are more likely to be underidentified
and/or fail to converge, and error estimates may be unreliable.

Some Questions
Sample Size
In relation to parameters, what sample size does Kline recommend? He
recommends 10 times as many cases as parameters (or ideally 20 times).
He states that 5 times or less is insufficient for significance testing of model
effects. Rex B. Kline
To have confidence in the goodness of fit test, a sample size of 100 to 200
is recommended (Hoyle 1995). In general a model should contain 10 to 20
times as many observations as variables (Mitchell 1993).
Sample Size. As with factor analysis, you should have lots of data when
evaluating a SEM. As usual, there are several rules of thumb. For a simple
model, 200 cases might be adequate. When relationships among
components of the model are strong, 10 cases per estimated parameter may
be adequate. Karl L. Wuensch Dept. of Psychology, East Carolina
University, Greenville, NC USA November, 2009


- SEM
Principles and Practice of Structural
Equation Modeling
Rex B. Kline NY: Guilford Press, 1998.

http://student.sut.ac.th/buratin/index.php/using-mplus/49-rex-b-kline