1

SEM : STRUCTURAL EQUATION MODEL

PENDAHULUAN
SEM adalah teknik yang dapat digunakan untuk menentukan dan
estimasi model hubungan linier antara variabel. Variabel dalam model SEM ini
dapat mencakup variabel yang dapat diukur (MV) dan variabel laten (LV). LV
adalah konstruks- hipotetis yang tidak dapat diukur secara langsung. Dalam SEM
setiap konstruks diwakili oleh beberapa MV yang berfungsi sebagai indikator
(atau faktor, atau ukuran). Sebuah model SEM, pada dasarnya adalah pola
hipotetik hubungan linear directional dan nondirectional antara himpunan MV dan
LV. Hubungan directional mengisyaratkan semacam pengaruh directional dari
satu variabel terhadap variable lainnya. Hubungan non-directional adalah
korelasional dan menyiratkan tidak-adanya arah pengaruhnya. Dalam bentuk
yang paling umum dari SEM, tujuan dari model ini adalah untuk menjelaskan
variasi dan kovariasi dari MV.
Pemodelan persamaan structural (SEM-Structural Equation Modelling)
adalah, teknik pemodelan statistik yang bersifat general, terutama linear, dan
cross-sectional. Analisis faktor, analisis jalur dan analisis regresi, semua
merupakan kasus khusus dari SEM.
SEM lebih cenderung sebagai teknik konfirmasi, bukan teknik eksplorasi.
Seorang peneliti lebih cenderung menggunakan SEM untuk menentukan apakah
model hipotetiknya valid, daripada menggunakan SEM untuk "menemukan"
model yang cocok. Meskipun demikian, analisis SEM seringkali melibatkan
elemen-elemen eksplorasi tertentu.
Dalam analisis SEM, biasanya minatnya berfokus pada konstruks laten (
variabel psikologis yang abstrak) seperti "kecerdasan" atau "sikap terhadap
merek", bukan berfokus pada variabel manifes yang digunakan untuk mengukur
konstruks-laten ini. Pengukuran memang sangat sulit dan rawan kesalahan. Para
pengguna SEM berusaha untuk men-derivasi estimasi-estimasi tanpa-bias untuk
hubungan antara konstruk-laten. Untuk tujuan ini, SEM memungkinkan ukuran
ganda untuk suatu konstruks-laten tunggal.
Sebuah model persamaan struktural menyiratkan adanya struktur matriks
kovarians dari ukuran-ukuran (sehingga sering disebut sebagai "analisis
kovarians struktur"). Setelah parameter model diestimasi, model yang dihasilkan
(matriks kovarians) kemudian dibandingkan dengan matriks kovariansi empiris
atau matriks kovarians berdasarkan data. Jika dua matriks ini saling konsisten
satu sama lain, maka model persamaan struktural dapat dianggap sebagai
penjelasan yang masuk akal bagi hubungan-hubungan di antara ukuran-ukuran.
SEM merupakan bidang-kajian yang relatif baru dibandingkan dengan
analisis regresi dan analisis faktor, dipublikasi pada akhir tahun 1960. Dengan
demikian, metodologi masih sedang berkembang, dan bahkan konsep dasarnya
masih menantang untuk direvisi.
Dalam SEM, variabel kunci biasanya berupa "konstruks-laten", yaitu
konsep-konsep psikologis yang abstrak, seperti "kecerdasan" atau "sikap".
Perilaku konstruk-laten ini hanya dapat diamati secara tidak langsung, dan tidak
sempurna, melalui pengaruhnya terhadap variabel-manifes. Suatu Model SEM
dapat mencakup dua jenis konstruks-laten, yaitu eksogen dan endogen. Dalam
sistem yang paling tradisional, konstruks-eksogen ditunjukkan dengan symbol
karakterhuruf Yunani "ksi" (, di sebelah kiri) dan konstruks-endogen ditunjukkan
dengan karakter Yunani "eta" (, di sebelah kanan). Kedua jenis konstruks ini

dibedakan atas dasar apakah atau mereka merupakan variabel dependen pada
suatu persamaan dalam sistem-persamaan diwakili oleh model. Konstrukseksogen adalah variabel independen dalam semua persamaannya, sedangkan
konstruks-endogen adalah variabel dependen dalam setidaknya satu
persamaan, meskipun mereka mungkin menjadi variabel independen dalam
persamaan lain dalam sistem-persamaan. Dalam Model-grafis, setiap konstruksendogen merupakan target dari minimal satu anak panah, sedangkan konstrukseksogen hanya ditargetkan oleh anak-panah berkepala dua.
Model SEM menggabungkan beberapa pendekatan atau kerangka kerja
yang berbeda-beda untuk mewakili model ini. Suatu model persamaan struktural
yang umum dilukiskan dengan tiga persamaan matriks. Akan tetapi, dalam
prakteknya, model SEM paling sering disajikan dalam bentuk grafik (Bagan).

Gambar 1. Contoh Representasi Model SEM.

Dalam SEM, model struktural mencakup hubungan di antara konstruklaten. Hubungan ini terutama bersifat linear, meskipun dimungkinkan untuk
masuknya hubungan nonlinier. Dalam bentuk diagram, anak-panah berkepala
satu merupakan hubungan regresi, sedangkan anak panah berkepala dua
mewakili hubungan korelasional.
Parameter yang mewakili hubungan regresi antara konstruks-laten
biasanya diberi label dengan karakter Yunani "gamma" () untuk regresi
konstruks-endogen pada konstruks-eksogen; atau dengan karakter Yunani "beta"
() untuk regresi dari satu konstruks-endogen pada konstruks-endogen lainnya.
Biasanya dalam suatu Model SEM, konstruks-eksogen diperbolehkan untuk
covary bebas. Parameter diberi label dengan karakter Yunani "phi" ()
merupakan kovarians ini. Kovarians ini berasal dari prediktor umum dari
konstruks-eksogen yang berada di luar model.
Para peneliti SEM berharap untuk dapat secara sempurna memprediksi
konstruks-tergantung, sehingga model biasanya mencakup istilah kesalahan
struktural, diberi label dengan karakter Yunani "zeta" (). Untuk mencapai
estimasi parameter yang konsisten, kesalahan ini diasumsikan tidak berkorelasi

dengan konstruk-eksogen dari model. Pelanggaran asumsi ini akibat dari

masalah prediktor masih diperbolehkan. Namun demikian, kesalahan struktural
dapat dimodelkan berkorelasi dengan kesalahan-struktural lainnya. Spesifikasi
tersebut menunjukkan bahwa konstruks-endogen yang terkait dengan istilahistilah error berbagi variasi yang tidak dijelaskan oleh hubungan prediktor dalam
model.
Peneliti SEM menggunakan variabel manifes (yaitu, ukuran aktual dan
skor actual) untuk melandasi model konstruks-latennya dengan data riil. Variabel
manifes yang terkait dengan konstruks-eksogen diberi label X, sedangkan yang
berhubungan dengan konstruks-endogen diberi label Y. Dalam SEM, masingmasing konstruks-laten biasanya dikaitkan dengan beberapa ukuran (indikator).
Peneliti SEM biasanya menghubungkan konstruks-laten dengan ukuran
(indicator)nya melalui model analisis faktor. Masing-masing konstruks-laten
dimodelkan sebagai faktor umum yang dilandasi oleh indikator-indikatornya.
Faktor "loading" ini, yang menghubungkan suatu konstruks dengan indikatornya
diberi label dengan karakter Yunani "lambda" (). Model persamaan struktural
dapat mencakup dua "matriks lambda yang berbeda, pada satu sisi adalah X
dan pada sisi lainnya adalah Y. Dalam aplikasi SEM, model pengukuran yang
paling lazim adalah model pengukuran congeneric, dimana masing-masing
indicator dikaitkan dengan satu konstruks-laten saja, dan semua kovariasi di
antara indicator merupakan konsekuensi dari hubungan antara indicator dan
konstruksnya.
Kadang-kadang, lebih masuk akal untuk model konstruks-laten sebagai
hasil atau konsekuensi dari indikator-indikatornya. Ini merupakan model indikator
kausal. Model pengukuran alternatif ini juga menjadi pusat Partial Least Squares,
metodologi yang terkait dengan SEM.
Pengguna SEM biasanya mengakui bahwa indikator-indikatornya tidak
sempurna, dan mereka berusaha untuk memodel ketidak-sempurnaan ini.
Dengan demikian, model SEM mencakup hal-hal yang mewakili kesalahan
pengukuran. Dalam konteks model pengukuran analisis faktor, istilah-istilah
kesalahan pengukuran yang unik atau faktor unik yang terkait dengan masingmasing indikator. Istilah kesalahan-pengukuran yang terkait dengan indicator X
diberi label dengan karakter Yunani "delta" (), sedangkan istilah yang terkait
dengan indikator Y diberi label dengan karakter "epsilon" (). Secara konseptual,
hampir setiap indicator memiliki kesalahan yang terkait. Dengan kata lain,
hampir setiap indicator menyertakan beberapa kesalahan.
Namun demikian, ketika suatu konstruks dikaitkan dengan indikator
tunggal, biasanya tidak mungkin (karena batas identifikasi) untuk memperkirakan
jumlah kesalahan pengukuran dalam model. Dalam kasus seperti ini, peneliti
harus menetapkan lebih dahulu besarnya kesalahan pengukuran sebelum
mencoba untuk memperkirakan parameter model. Dalam situasi seperti ini,
peneliti mungkin hanya mengasumsikan bahwa tidak ada kesalahan pengukuran.
Namun demikian, jika asumsi ini salah, maka estimasi parameter model yang
diperoleh akan bias.
Analisis factor eksploratori (EFA- Exploratory factor analysis) dan analisis
faktor konfirmatori (CFA- confirmatory factor analysis), Analisis jalur, dan Model
SEM memiliki sejarah panjang dalam berbagai penelitian ilmiah. Meskipun
sebagian besar CFA telah menggantikan EFA, namun konstruks
multidimensional CFA biasanya gagal untuk memenuhi standar pengukuran yang
baik: goodness of fit, pengukuran invarian, lack of differential item functioning,
dan well-differentiated factors,
dalam mendukung validitas diskriminan.
Sebagian dari masalah ini terjadi karena mengandalkan CFA terlalu ketat,

dimana setiap item hanya loading dengan satu indikator (faktor). Eksploratori
SEM (ESEM), mengintegrasikan secara menyeluruh keunggulan dari CFA / SEM
dan EFA-tradisional, menyediakan uji konfirmatori dari struktur faktor-priori,
hubungan antara faktor-faktor laten dengan uji-uji multigroup / multioccasion
penuh (struktur rata-rata) pengukuran invarian. Hal ini menggabungkan semua
kombinasi dari faktor CFA, faktor ESEM, kovariat, pengelompokan / MIMIC
(multiple-indicator multiple-cause) variabel, pertumbuhan-laten, dan struktur
kompleks, yang secara khusus memerlukan CFA / SEM.
Penggunaan berbagai metode pemodelan SEM dalam penelitian sosial
terus berkembang. Karena semakin banyaknya peneliti studi sosial berkenalan
dengan pemodelan persamaan struktural (SEM), telah terjadi peningkatan dalam
penerapan metode ini dalam jurnal-jurnal ilmu-ilmu sosial. Faktor-faktor yang
terkait dengan peningkatan penggunaan SEM dalam penelitian sosial adalah
ketersediaan perangkat lunak statistik, seperti LISREL, AMOS dan EQS, serta
perangkat lunak tambahan termasuk Mplus, SAS (Proc Calis), Mx dan R
(lavaan.org) (Okech, Kim dan Little, 2013).
Pemodelan persamaan struktural (SEM) memungkinkan peneliti menguji
hubungan antara variabel yang terukur dengan variable laten, serta menguji
kesamaan lintas-kelompok dan perbedaan antar variabel-laten (Brown, 2006).
Rigdon (2001) mendefinisikan SEM sebagai metode untuk mewakili,
memperkirakan dan menguji jaringan teoritis hubungan linier antara variabel
yang diamati dan variabel laten. SEM memiliki kemampuan untuk mengoreksi
kesalahan pengukuran dan memiliki asumsi lebih sedikit daripada metode klasik
lainnya (Little, 2013). Aplikasi lainnya dari SEM adalah dalam pengujian
perbedaan laten antara kelompok longitudinal, pemodelan efek kontekstual
dalam studi longitudinal antara kelompok (Little et al., 2007). Menurut Guo et al.
(2009), metode SEM memainkan peran penting dalam mengembangkan
pengetahuan untuk profesi sosial, karena mengintegrasikan pengukuran dengan
teori substantif.

ANALISIS SEM.
Structural equation modeling (SEM) is a general term used to
describe a family of statistical methods designed to test a conceptual or
theoretical model (Kaplan, 2007). Some common SEM methods include
confirmatory factor analysis, path analysis, and latent growth modeling
(Kline, 2011a). The term "structural equation model" most commonly refers
to a combination of two things: a "measurement model" that defines latent
variables using one or more observed variables, and a "structural
regression model" that links latent variables together ((Kaplan, 2007;
Kline, 2011b). The parts of a structural equation model are linked to one
another using a system of simultaneous regression equations (Kline,
2011c).
SEM is widely used in the social sciences because of its ability to
isolate observational error from measurement of latent variables ( Hancock,
2003). To provide a simple example, the concept of human intelligence
cannot be measured directly as one could measure height or weight.
Instead, psychologists develop theories of intelligence and write
measurement instruments with items (questions) designed to measure
intelligence according to their theory (Thorndike, 2007). They would then

use SEM to test their theory using data gathered from people who took
their intelligence test. With SEM, "intelligence" would be the latent variable
and the test items would be the observed variables.
A simplistic model suggesting that intelligence (as measured by five
questions) can predict academic performance (as measured by SAT, ACT,
and high school GPA) is shown below. In SEM diagrams, latent variables
are commonly shown as ovals and observed variables as rectangles. The
below diagram shows how error (e) influences each intelligence question
and the SAT, ACT, and GPA scores, but does not influence the latent
variables. SEM provides numerical estimates for each of the parameters
(arrows) in the model to indicate the strength of the relationships. Thus, in
addition to testing the overall theory, SEM therefore allows the researcher
to diagnose which observed variables are good indicators of the latent
variables (Hancock, 2003).

Gambar 2. A conceptual illustration of a structural equation model.

(http://en.wikipedia.org/wiki/Structural_equation_modeling).
Modern studies usually test much more specific models involving
several theories, for example, Jansen, Scherer, and Schroeders (2015)
studied how students' self-concept and self-efficacy affected educational
outcomes (Jansen, Scherer dan Schroeders, 2015). SEM is also used in the
sciences (Gillespie dan Perron, 2007), business (Markus, 2007), education
(Shelley, 2007), and many other fields.
SEM evolved in three different streams: (1) systems of equation
regression methods developed mainly at the Cowles Commission; (2)
iterative maximum likelihood algorithms for path analysis developed
mainly at the University of Uppsala by Karl Gustav Jreskog; and (3)
iterative canonical correlation fit algorithms for path analysis also
developed at the University of Uppsala by Hermann Wold. Much of this
development occurred at a time that automated computing was offering
substantial upgrades over the existing calculator and analogue computing
methods available, themselves products of the proliferation of office
equipment innovations in the late 19th century.
Loose and confusing terminology has obscured exactly what SEM
is doing with data. In particular, PLS-PA (the Lohmoller algorithm) is quite
commonly confused with partial least squares regression, which is
typically just called PLS. PLS regression tends to be useful with very large,
multicolinear datasets, and finds applications in spectroscopy. PLS-PA, in
contrast, is typically promoted as a method that works with small datasets

when other estimation approaches fail; though this contention, even in the
1970s, was known not to be true; e.g. see (Dhrymes, 1972, 1974;
Dhrymes & Erlat, 1972; Dhrymes et al., 1972; Gupta, 1969; Sobel, 1982)
Both LISREL and PLS-PA were conceived as iterative computer
algorithms, with an emphasis from the start on creating an accessible
graphical and data entry interface and extension of Wrights (1921) path
analysis. Early Cowles Commission work on simultaneous equations
estimation centered on Koopman and Hoods (1953) algorithms from the
economics of transportation and optimal routing, with maximum likelihood
estimation, and closed form algebraic calculations, as iterative solution
search techniques were limited in the days before computers. Anderson
and Rubin (1949, 1950) developed the limited information maximum
likelihood estimator for the parameters of a single structural equation,
which indirectly included the two-stage least squares estimator and its
asymptotic distribution (Anderson, 2005) and Farebrother (1999). Twostage least squares was originally proposed as a method of estimating the
parameters of a single structural equation in a system of linear
simultaneous equations, being introduced by Theil (1953a, 1953b, 1961)
and more or less independently by Basmann (1957) and Sargan (1958).
Andersons limited information maximum likelihood estimation was
eventually implemented in a computer search algorithm, where it
competed with other iterative SEM algorithms. Of these, two-stage least
squares was by far the most widely used method in the 1960s and the
early 1970s.
LISREL and PLS path modeling approaches were championed at
the Cowles Commission mainly by Nobelist Trygve Haavelmo (1943),
while the underlying assumptions of LISREL and PLS were challenged by
statisticians such as Freedman (1987) who objected to their failure to
distinguish among causal assumptions, statistical implications, and policy
claims has been one of the main reasons for the suspicion and confusion
surrounding quantitative methods in the social sciences (see also Wolds
(1987) response). Haavelmos path analysis never gained a large following
among U.S. econometricians, but was successful in influencing a
generation of Haavelmos fellow Scandinavian statisticians, including
Hermann Wold, Karl Jreskog, and Claes Fornell. Fornell introduced
LISREL and PLS techniques to many of his Michigan colleagues through
influential papers in accounting (Fornell and Larker 1981), and information
systems (Davis, et al., 1989). Dhrymes (1971; Dhrymes, et al. 1974)
provided evidence that PLS estimates asymptotically approached those of
two-stage least squares with exactly identified equations. This point is
more of academic importance than practical, because most empirical
studies overidentify. But in one sense, all of the limited information
methods (OLS excluded) yield similar results.
Advances in computers and the exponential increase in data
storage have created many new opportunities to apply structural equation
methods in the computer-intensive analysis of large datasets in complex,
unstructured problems. The most popular solution techniques fall into
three classes of algorithms: (1) ordinary least squares algorithms applied

independently to each path, such as applied in the so-called PLS path

analysis packages which may estimate with OLS or PLSR; (2) covariance
analysis algorithms evolving from seminal work by Wold and his student
Karl Jreskog implemented in LISREL, AMOS, and EQS; and (3)
simultaneous equations regression algorithms developed at the Cowles
Commission by Tjalling Koopmans. The popularity of SEM path analysis
methods in the social sciences reects a more holistic, and less blatantly
causal, interpretation of many real world phenomena especially in
psychology and social interaction than may be adopted in the natural
sciences. Direction in the directed network models of SEM arises from
presumed cause-effect assumptions made about reality. Social
interactions and artifacts are often epiphenomena secondary
phenomena that are difficult to directly link to causal factors. An example
of a physiological epiphenomenon is, for example, time to complete a 100
meter sprint. I may be able to improve my sprint speed from 12 seconds to
11 seconds but I will have difficulty attributing that improvement to any
direct causal factors, like diet, attitude, weather, etc. The 1 second
improvement in sprint time is an epiphenomenon the holistic product of
interaction of many individual factors.
General Approach to SEM
Structural equation modeling, or SEM, is a very general, chiefly
linear, chiefly cross-sectional statistical modeling technique. Factor
analysis, path analysis and regression all represent special cases of SEM.
SEM is 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.
In SEM, interest usually focuses on latent constructs--abstract
psychological variables like "intelligence" or "attitude toward the brand"-rather than on the manifest variables used to measure these constructs.
Measurement is recognized as difficult and error-prone. By explicitly
modeling measurement error, SEM users seek to derive unbiased
estimates for the relations between latent constructs. To this end, SEM
allows multiple measures to be associated with a single latent construct.
A structural equation model implies a structure of the covariance
matrix of the measures (hence an alternative name for this field, "analysis
of covariance structures"). Once the model's parameters have been
estimated, the resulting model-implied covariance matrix can then be
compared to an empirical or data-based covariance matrix. If the two
matrices are consistent with one another, then the structural equation
model can be considered a plausible explanation for relations between the
measures.
Compared to regression and factor analysis, SEM is a relatively
young field, having its roots in papers that appeared only in the late 1960s.
As such, the methodology is still developing, and even fundamental
concepts are subject to challenge and revision. This rapid change is a

source of excitement for some researchers and a source of frustration for

others.
Although each technique in the SEM family is different, the following
aspects are common to many SEM methods.
Model specification
Two main components of models are distinguished in SEM: the
structural model showing potential causal dependencies between
endogenous and exogenous variables, and the measurement model
showing the relations between latent variables and their indicators.
Exploratory and Confirmatory factor analysis models, for example, contain
only the measurement part, while path diagrams can be viewed as SEMs
that contain only the structural part.
In specifying pathways in a model, the modeler can posit two types
of relationships: (1) free pathways, in which hypothesized causal (in fact
counterfactual) relationships between variables are tested, and therefore
are left 'free' to vary, and (2) relationships between variables that already
have an estimated relationship, usually based on previous studies, which
are 'fixed' in the model.
A modeler will often specify a set of theoretically plausible models in
order to assess whether the model proposed is the best of the set of
possible models. Not only must the modeler account for the theoretical
reasons for building the model as it is, but the modeler must also take into
account the number of data points and the number of parameters that the
model must estimate to identify the model. An identified model is a model
where a specific parameter value uniquely identifies the model, and no
other equivalent formulation can be given by a different parameter value. A
data point is a variable with observed scores, like a variable containing the
scores on a question or the number of times respondents buy a car. The
parameter is the value of interest, which might be a regression coefficient
between the exogenous and the endogenous variable or the factor loading
(regression coefficient between an indicator and its factor). If there are
fewer data points than the number of estimated parameters, the resulting
model is "unidentified", since there are too few reference points to account
for all the variance in the model. The solution is to constrain one of the
paths to zero, which means that it is no longer part of the model.
Structural Model
In SEM, the structural model includes the relationships among the
latent constructs. These relationships are chiefly linear, although flexible
extensions to the basic SEM system allow for the inclusion of nonlinear
relations, as well. In the diagram, one-headed arrows represent regression
relationships, while two-headed arrows represent correlational relations-that is, shared variation that is not explained within the model.
Structural equation modeling incorporates several different
approaches or frameworks to representing these models. In one well-

known framework (popularized by Karl Jreskog, University of Uppsala),

the general structural equation model can be represented by three matrix
equations:
However, in applied work, structural equation models are most
often represented graphically. Here is a graphical example of a structural
equation model:

10

Gambar
. Model-model konseptual yang disajikan dlaam bentuk sederhana
(http://www.nwrc.usgs.gov/SEM/SEM.1.2-SEM%20Essentials%20-%20Model
%20Anatomy%20-%20version%201.0.pdf).

Gambar
. Variabel eksogen (independent) dan endogen
(dependent)
(http://www.nwrc.usgs.gov/SEM/SEM.1.2-SEM
%20Essentials%20-%20Model%20Anatomy%20-%20version
%201.0.pdf).

11

Gambar

.
Hubungan
dalam
jalur-jalur.
(http://www.nwrc.usgs.gov/SEM/SEM.1.2-SEM%20Essentials%20%20Model%20Anatomy%20-%20version%201.0.pdf).
Each pathway tracing a route from a predictor to a
response represents a distinct mechanism. In this slide I
am trying to avoid a common misrepresentation related to
the effect terminology (e.g., direct effect). That
terminology causes lots of problems in presentations
because it seems to confuse results with conclusions. It is
helpful to be explicit that these are only effects in the
model, not in the real system. Two things are meant here
by this statement.
(1) Caution must be taken when interpreting a direct effect
in an SEM because in reality this direct effect is likely
mediated by an unmeasured variable. In real systems,
we assume that effects can be infinitely decomposed
and ultimately operate at the quantum level. Thus, all
true effects are indirect but since we are working with
observed variables in our models, the effects actually
specified in the models are simplified down to direct
and indirect effects (in the model).
(2) The conclusion that there is an effect depends on
some causal assumptions that are not actually tested
with the data. Thus, effect in the model clarifies that
we recognize that critical assumption.

12

Gambar . Beberapa macam arsitektur model SEM.

Here is some terminology related to model architecture.
Saturated models have links between all the variables (more
strictly true, we have no model degrees of freedom).
Saturated models represent a special class of model because
they allow for everything to add up, meaning we can
completely recover the observed matrix of covariances when
our model is saturated. Unsaturated models have testable
implications, however. Specifically, for Model A, the model
hypothesizes that the indirect effect of x1 on y2 equals the
observed covariance between those variables. This may not
be empirically true, which would imply that there is some
other connection between x1 and y2, such as in Model B.
Model C is recursive in that causation flows without loops. Model C also
allows for x1 and x2 to be correlated for some unspecified reason.
Nonrecursive models, such as in Model D, have more complex causal
relationships. Here it is shown the case of a reciprocal interaction, which is a
type of causal loop. To understand such models, we really need to consider
the temporal dynamics that underlie the model, along with the assumptions
required for collapsing time.
Latent Constructs
In structural equation modeling, the key variables of interest are
usually "latent constructs"--abstract psychological concepts such as

13

"intelligence" or "attitude." We can observe the behavior of latent variables

only indirectly, and imperfectly, through their effects on manifest variables.
A structural equation model may include two types of latent
constructs--exogenous and endogenous. In the most traditional system,
exogenous constructs are indicated by the Greek character "ksi" (at left)
(i) and endogenous constructs are indicated by the Greek character "eta"
(at right) ().
These two types of constructs are distinguished on the basis of
whether or not they are dependent variables in any equation in the system
of equations represented by the model. Exogenous constructs are
independent variables in all equations in which they appear, while
endogenous constructs are dependent variables in at least one equation-although they may be independent variables in other equations in the
system. In graphical terms, each endogenous construct is the target of at
least one one-headed arrow, while exogenous constructs are only targeted
by two-headed arrows.
Parameters representing regression relations between latent
constructs are typically labeled with the Greek character "gamma" ( ) for
the regression of an endogenous construct on an exogenous construct, or
with the Greek character "beta" () for the regression of one endogenous
construct on another endogenous construct.
Typically in SEM, exogenous constructs are allowed to covary
freely. Parameters labeled with the Greek character "phi" () represent
these covariances. This covariance comes from common predictors of the
exogenous constructs which lie outside the model under consideration.
Structural Error
Few SEM researchers expect to perfectly predict their dependent
constructs, so model typically include a structural error term, labeled with
the Greek character "zeta" ( ). To achieve consistent parameter
estimation, these error terms are assumed to be uncorrelated with the
model's exogenous constructs. (Violations of this assumption come about
as a result of the excluded predictor problem.) However, structural error
terms may be modeled as being correlated with other structural error
terms. Such a specification indicates that the endogenous constructs
associated with those error terms share common variation that is not
explained by predictor relations in the model.
Manifest Variables
SEM researchers use manifest variables--that is, actual measures
and scores--to ground their latent construct models with real data.
Manifest variables associated with exogenous constructs are labeled X,
while those associated with endogenous constructs are labeled Y.
Otherwise, there is no fundamental distinction between these measures,
and a measure that is labeled X in one model may be labeled Y in another.
Measurement Model

14

In SEM, each latent construct is usually associated with multiple

measures. SEM researchers most commonly link the latent constructs to
their measures through a factor analytic measurement model. That is,
each latent construct is modeled as a common factor underlying the
associated measures. These "loadings" linking constructs to measures are
labeled with the Greek character "lambda" (at left). Structural equation
models can include two separate "lambda matrices, one on the X side and
one on the Y side. In SEM applications, the most common measurement
model is the congeneric measurement model, where each measure is
associated with only one latent construct, and all covariation between
measures is a consequence of the relations between measures and
constructs.
(Sometimes, however, it makes more sense to model a latent
construct as the result or consequence of its measures. This is the causal
indicators model. This alternative measurement model is also central to
Partial Least Squares, a methodology related to SEM.)
Measurement Error
SEM users typically recognize that their measures are imperfect,
and they attempt to model this imperfection. Thus, structural equation
models include terms representing measurement error. In the context of
the factor analytic measurement model, these measurement error terms
are uniquenesses or unique factors associated with each measure.
Measurement error terms associated with X measures are labeled with the
Greek character "delta" (at left) while terms associated with Y measures
are labeled with "epsilon" (at right). Conceptually, almost every measure
has an associated error term. In other words, almost every measure is
acknowledged to include some error.
However, when a construct is associated with only a single
measure, it is usually impossible (due to the limits of identification) to
estimate the amount of measurement error within the model. In such
cases, the researcher must prespecify the amount of measurement error
before attempting to estimate model parameters. In this situation,
researchers may be tempted to simply assume that there is no
measurement error. However, if this assumption is false, then model
parameter estimates will be biased.
Estimation of free parameters
Parameter estimation is done by comparing the actual covariance
matrices representing the relationships between variables and the
estimated covariance matrices of the best fitting model. This is obtained
through numerical maximization of a fit criterion as provided by maximum
likelihood estimation, quasi-maximum likelihood estimation, weighted least
squares or asymptotically distribution-free methods. This is often
accomplished by using a specialized SEM analysis program, of which
several exist.
Assessment of model and model fit

15

Having estimated a model, analysts will want to interpret the model.

Estimated paths may be tabulated and/or presented graphically as a path
model. The impact of variables is assessed using path tracing rules (see
path analysis).
It is important to examine the "fit" of an estimated model to
determine how well it models the data. This is a basic task in SEM
modeling: forming the basis for accepting or rejecting models and, more
usually, accepting one competing model over another. The output of SEM
programs includes matrices of the estimated relationships between
variables in the model. Assessment of fit essentially calculates how similar
the predicted data are to matrices containing the relationships in the actual
data.
Formal statistical tests and fit indices have been developed for
these purposes. Individual parameters of the model can also be examined
within the estimated model in order to see how well the proposed model
fits the driving theory. Most, though not all, estimation methods make such
tests of the model possible.
Of course as in all statistical hypothesis tests, SEM model tests are
based on the assumption that the correct and complete relevant data have
been modeled. In the SEM literature, discussion of fit has led to a variety
of different recommendations on the precise application of the various fit
indices and hypothesis tests.
There are differing approaches to assessing fit. Traditional
approaches to modeling start from a null hypothesis, rewarding more
parsimonious models (i.e. those with fewer free parameters), to others
such as AIC that focus on how little the fitted values deviate from a
saturated model (i.e. how well they reproduce the measured values),
taking into account the number of free parameters used. Because different
measures of fit capture different elements of the fit of the model, it is
appropriate to report a selection of different fit measures. Guidelines (i.e.,
"cutoff scores") for interpreting fit measures, including the ones listed
below, are the subject of much debate among SEM researchers
(MacCallum dan Austin, 2000).
Some of the more commonly used measures of fit include:
1. Chi-Squared A fundamental measure of fit used in the
calculation of many other fit measures. Conceptually it is a
function of the sample size and the difference between the
observed covariance matrix and the model covariance matrix.
2. Akaike information criterion (AIC)
A test of relative model fit: The preferred model is the one with
the lowest AIC value.

where k is the number of parameters in the statistical model,

and L is the maximized value of the likelihood of the model.
3. Root Mean Square Error of Approximation (RMSEA)

16

Fit index where a value of zero indicates the best fit (Kline,
2011). While the guideline for determining a "close fit" using
RMSEA is highly contested (Kline, 2011), most researchers
concur that an RMSEA of .1 or more indicates poor fit (Hu dan
Bentler, 1999).
4. Standardized Root Mean Residual (SRMR)
The SRMR is a popular absolute fit indicator. Hu and Bentler
(1999) suggested .08 or smaller as a guideline for good fit (Hu
dan Bentler, 1999).
5. Comparative Fit Index (CFI)
In examining baseline comparisons, the CFI depends in large
part on the average size of the correlations in the data. If the
average correlation between variables is not high, then the CFI
will not be very high. A CFI value of .95 or higher is desirable
(Hu dan Bentler, 1999).
For each measure of fit, a decision as to what represents a goodenough fit between the model and the data must reflect other contextual
factors such as sample size, the ratio of indicators to factors, and the
overall complexity of the model. For example, very large samples make
the Chi-squared test overly sensitive and more likely to indicate a lack of
model-data fit (Kline, 2011).

Gambar . Model SEM dan kaitannya dnegan korelasi

For this presentation, it is helpful to adopt a linear Gaussian
approximation. One thing this allows us to do is to summarize our

17

data in the form of a covariance matrix, which when standardized, is

a correlation matrix. Pearl (2013) uses the same approach to
connect historical SEM to his nonparametric (Pearl, J. 2013. Linear
models: a useful microscope for causal analysis. Journal of Causal
Inference. 1:155-170).

Gambar
. Koefisien Jalur Kaidah 1. It is useful to know that
exogenous correlations/covariance are not computed by SEM packages. Rather, we
can pull them directly from a covariance matrix. This practice is predicated on the
first rule of path coefficients.

18

Gambar . Degresi tentang korelasi dalam Model SEM. Before moving

on, we should realize that correlations in models actually imply some causal process
involving hidden variables. Essentially these are latent processes of some type, but we
generally dont go so far as to represent them that way, UNLESS, we are doing latent
variable modeling (see separate modules for that).

19

Gambar . Koefisien Jalur Kaidah 2. When variables are connected by a

single causal pathway, coefficients are simple. This means they will correspond
with the observed correlations.

20

Gambar . Koefisien Jalur- Kaidah 3. We multiply coefficients along

compound pathways to get the expected net relationship. Since the second rule tells
us that for this model the path coefficients should be the observed correlations, we
can then compute the indirect effect of x1 on y2, which also represents the implied
correlation between these two variables. This then allows us to introduce and
illustrate a critically-important concept conditional independence. If the model
holds here, then once we know the values for y1, we dont need the values of x1 to
predict y2. This is why we say that x1 and y2 are independent, once we condition on y1.

Gambar . Koefisien Jalur Kaidah 4. Continuing, when two variables are

connected by more than one causal pathway, parameters/coefficients are not simple
anymore. Note that we cannot pick the values out of the covariance matrix when
there are partial effects in the model.

21

Gambar
. Koefisien jalur Kaidah 4. Returning to those pesky exogenous
correlations, they have causal interpretations, but not clean ones. There are two causal
connections between xs and y, but the effects on y are shared across the xs in this
model.

22

Gambar . Koefisien Jalur Kaidah 5. A key part of our model is the

influences of other factors or errors of prediction. The estimates for these quantities
are computed from the other information. They can be represented in different ways
and these are important to know for a variety of reasons. First, we can express the
variance explained for an endogenous variable by giving the R-square. This
represents 1-error variance in standardized terms. Second, we can actually present
the quantity of error variances or zetas, either in raw or standardized units. Third, if
we wish to treat error variables like true causal influences, then we might use path
coefficients for their effects. These = the square roots of the error variances (e.g.,
sqrt(0.84) = 0.92) or alternatively, the square roots of 1-R-square.

23

Gambar . Koefisien Jalur Kaidah 6. Considering models with joint

responses are quite instructive for multiple reasons. Here our focus is on estimating
residual correlations/covariances. The basic point is that the implied correlation
between two joint responses is the product of the path coefficients connecting them.
In this case, the observed correlation is 0.60 while the model-implied correlation is
0.20. This is a pretty big difference, implying a residual correlation between zeta1 and
zeta2.

24

Gambar . Koefisien Jalur Kaidah 6. When the observed correlation is

non-trivially different from that implied by their joint responses to x1, in this case, it
implies some other (latent) factor. This additional connection is expressed in the form
of a residual correlation among errors. Our goal here is to compute the residual
correlation by hand.
Step 1: Compute the R-squares for the two responses (0.16 and 0.25 in this case.
Step 2: Compute the path coefficients for the error influences on the variables. Here
we subtract the R-squares from 1.0 to get the error variances, then take the square
roots of those to get the path coefficients (0.92 and 0.86 in this case).
Step 3: The total correlation between y1 and y2 will be the sum of the two compound
pathways connecting them.
Path 1: 0.50*0.40 = 0.2
Path 2: 0.92*0.86*? = the difference
This leads us to an estimate of 0.50 for the standardized error correlation.

Gambar . Koefisien Jalur Kaidah 7. Total effect is an important concept.

On the next slide we will contrast total effect with total correlation. For the
moment, we simply consider that the total effect represents the causal influence of a
predictor on a response when all mediating variables are allowed to change in
response to changes in the predictor. In this case, lets consider the total effect of x1
on y2. There is a direct component = 0.15. There is also one indirect path through y1
whose effect is 0.64*0.27. So, the total effect is the sum of the paths, or 0.15 +
(0.64*0.27) = 0.32.

25

Gambar . Koefisien Jalur Kaidah 8. We already introduced this basic

concept, but the total correlation between two variables represents the sum of all
pathways connecting them, including causal and non-causal paths. Here, the
correlation between x1 and x2 is considered a non-causal path because we have not
included in our model a variable to explain that effect. The computations for partial
effects are defined as being those that will recover the net correlations observed in the
data.

26

Gambar . Efek Supresi. Simpsons paradox was the observation that adding
variables changes the partial effects of other variables, often in radical ways. Here we
illustrate the common case where the observed correlation is very different from the
calculated causal connection. There is no real surprise here except that we often try to
understand causal relations between variables using bivariate patterns when we know
that is the wrong way to approach the problem.
Model modification
The model may need to be modified in order to improve the fit,
thereby estimating the most likely relationships between variables. Many
programs provide modification indices which may guide minor
modifications. Modification indices report the change in that result from
freeing fixed parameters: usually, therefore adding a path to a model which
is currently set to zero. Modifications that improve model fit may be
flagged as potential changes that can be made to the model. Modifications
to a model, especially the structural model, are changes to the theory
claimed to be true. Modifications therefore must make sense in terms of
the theory being tested, or be acknowledged as limitations of that theory.
Changes to measurement model are effectively claims that the items/data
are impure indicators of the latent variables specified by theory (Loehlin,
2004).
Models should not be led by MI, as Maccallum (1986)
demonstrated: "even under favorable conditions, models arising from
specification searches must be viewed with caution." (MacCallum, 1986).
Sample size and power
While researchers agree that large sample sizes are required to
provide sufficient statistical power and precise estimates using SEM, there
is no general consensus on the appropriate method for determining
adequate sample size (Westland, 2010). Generally, the considerations for
determining sample size include the number of observations per
parameter, the number of observations required for fit indexes to perform
adequately, and the number of observations per degree of freedom
(Quintana dan Maxwell, 1999). Researchers have proposed guidelines
based on simulation studies (Chou & Bentler, 1995), professional
experience (Bentler and Chou, 1987), and mathematical formulas
(MacCallum, Browne, and Sugawara, 1996; Westland, 2010).
Sample size requirements to achieve a particular significance an
power in SEM hypothesis testing are similar for the same model when any
of the three algorithms (PLS-PA, LISREL or systems of regression
equations) are used for testing.
Interpretation and communication
The set of models are then interpreted so that claims about the
constructs can be made, based on the best fitting model.
Caution should always be taken when making claims of causality
even when experimentation or time-ordered studies have been done. The

27

term causal model must be understood to mean: "a model that conveys
causal assumptions," not necessarily a model that produces validated
causal conclusions. Collecting data at multiple time points and using an
experimental or quasi-experimental design can help rule out certain rival
hypotheses but even a randomized experiment cannot rule out all such
threats to causal inference. Good fit by a model consistent with one causal
hypothesis invariably entails equally good fit by another model consistent
with an opposing causal hypothesis. No research design, no matter how
clever, can help distinguish such rival hypotheses, save for interventional
experiments (Judea, 2000).
As in any science, subsequent replication and perhaps modification
will proceed from the initial finding.

Gambar . Bagaimana interpretasi koefisien jalur. W e have been talking

about coefficients, but how should we think about them? This is actually a subtle and
complex topic and I plan to cover it in more depth in a separate module on
Interpreting Path Coefficients and Their Generalizations.
At the most basic level, the coefficients shown are predictions. They predict, for
example, the effect on y2 if we were to vary y1 while holding the value of x1
constant.
(http://www.nwrc.usgs.gov/SEM/SEM.1.6-SEM%20Essentials%20-%20Interpreting
%20Coefficients%20-%20version%201.0.pdf)

28

Gambar
. Interpretasi koefisien jalur yang masih mentah (belum
dibakukan). The direct interpretation of raw coefficients is pretty straightforward,
at least in simple terms.

Gambar
. Interpretasi koefisien jalur yang dibakukan. The interpretation
of conventional standardized coefficients is more complex. They express the
predicted changes in terms of standard deviation units. Note that the units for the raw
coefficient in this case are are:

29

changes y2 as a function of changes in x1 ; and when multiplied by the ratio of

standard deviations for x1/y2 causes the raw units to cancel out of the standardized
coefficient.

Gambar . Koefisien jalur yang dibakukan, diinterpretasikan dalam

konteks sampel.
The assumptions that are associated with a clean interpretation of standardized
coefficients have been questioned since Tukey raised this issue in 1954. Still today,
while scientists often use standardized coefficients, many statisticians object to their
use because they are sample specific and not generalizable. Be aware that there is
another related body of discussion that has gone on about the meaning of
importance, which I am ignoring here for simplicity.

30

Gambar
. Kisaran yang relevan dari standarisasi. Relevant range
coefficients will often be similar to regular standardized coefficients that are based on
standard deviations. Perhaps the biggest advantage to using a relevant range approach
is when comparing groups. When comparing coefficients across groups. regular
standardization requires equal variance across groups, which is almost never the case.
However, we may be able to establish a set of relevant ranges that apply across
groups, thereby making the coefficients comparable across groups.

31

Aplikasi Model SEM

1.

Ekosistem Danau
Belovsky, G. E., D. Stephens, C. Perschon, P. Birdsey, D. Paul,
D. Naftz, R. Baskin, C. Larson, C. Mellison, J. Luft, R. Mosley,
H. Mahon, J. Van Leeuwen, and D. V. Allen. 2011. The Great
Salt Lake Ecosystem (Utah, USA): long term data and a
structural
equation
approach.
Ecosphere
2(3):art33.
doi:10.1890/ES10-00091.1.

Great Salt Lake (Utah, USA) is one of the worlds largest hypersaline
lakes, supporting many of the western U.S.s migratory waterbirds. This
unique ecosystem is threatened, but it and other large hypersaline lakes
are not well understood. The ecosystem consists of two weakly linked food
webs: one phytoplankton-based, the other organic particle/benthic algaebased. Seventeen years of data on the phytoplankton-based food web are
presented: abundances of nutrients (N and P), phytoplankton
(Chlorophyta, Bacillariophyta, Cyanophyta), brine shrimp (Artemia
franciscana), corixids (Trichocorixa verticalis), and Eared Grebes
(Podiceps nigricollis). Abundances of less common species, as well as
brine fly larvae (Ephydra cinerea and hians) from the organic
particle/benthic algae-based food web are also presented. Abiotic
parameters were monitored: lake elevation, temperature, salinity, PAR,
light penetration, and DO.We use these data to test hypotheses about the
phytoplankton-based food web and its weak linkage with the organic
particle/benthic algae-based food web via structural equation modeling.
Counter to common perceptions, the phytoplankton-based food web is not
limited by high salinity, but principally through phytoplankton production,
which is limited by N and grazing by brine shrimp. Annual N abundance is
highly variable and depends on lake volume, complex mixing given
thermo- and chemo-clines, and recycling by brine shrimp. Brine shrimp are
food-limited, and predation by corixids and Eared Grebes does not
depress their numbers. Eared Grebe numbers appear to be limited by
brine shrimp abundance. Finally, there is little interaction of brine fly larvae
with brine shrimp through competition, or with corixids or grebes through
predation, indicating that the lakes two food webs are weakly connected.
Results are used to examine some general concepts regarding food web
structure and dynamics, as well as the lakes future given expected
anthropogenic impacts.

32

Gambar 1. Our hypothesized simple food web diagrams for the Great Salt
Lake South Arm are presented. Two food webs are proposed: a
phytoplankton-based web (right of dashed line) and an organic
particle/benthic algae web (left of dashed line). Arrow thickness represents
suspected relative interaction strengths. The phytoplanktonbased web was
examined here.

33

Gambar 2. The phytoplankton-based food web in the Great Salt Lake

South Arm is summarized for periods when brine shrimp are present
(MarchNovember) and absent (DecemberFebruary). Arrow size reflects
interaction strength, with blue arrows indicating indirect interactions and
black arrows indicating direct interactions. Beside the arrow is its partial
correlation coefficient and each factors r2 is in parentheses.

2. Manajemen Ekosistem Hutan

34

Figure 4: The fitted SR model with unstandardized parameter estimates. The

observed variables and the error arrows of endogenous latent variables
are omitted to simplify the diagram. Estimated unstandardized path
coefficients (direct effects) are placed alongside each path. The squared
multiple correlations ( ) are included for each endogenous latent
variable. The thickness of a single arrowhead corresponds to the strength
of its estimates (based on its standardized estimate). The dashed single
arrowheads depict non-significant paths. The double arrowheads depict
the error covariance between the errors of fine litter and coarse litter
latent variables (Sumber: Tzeng Yih Lam, and Douglas A. Maguire.
2012. Structural Equation Modeling: Theory and Applications in Forest
Management. International Journal of Forestry Research, Vol 2012, 16p.)
(http://www.hindawi.com/journals/ijfr/2012/263953/.

35

3.. Manajemen Ekosistem Seagrass

Tjisse van der Heide, Egbert H. van Nes, Marieke M. van
Katwijk, Han Olff, and Alfons J. P. Smolders. 2011. Positive
Feedbacks in Seagrass Ecosystems Evidence from LargeScale Empirical Data. PLoS One. 2011; 6(1): e16504.
Positive feedbacks cause a nonlinear response of ecosystems to
environmental change and may even cause bistability. Even though the
importance of feedback mechanisms has been demonstrated for many types of
ecosystems, their identification and quantification is still difficult. Here, we
investigated whether positive feedbacks between seagrasses and light conditions
are likely in seagrass ecosystems dominated by the temperate seagrass Zostera
marina. We applied a combination of multiple linear regression and structural
equation modeling (SEM) on a dataset containing 83 sites scattered across
Western Europe. Results confirmed that a positive feedback between sediment
conditions, light conditions and seagrass density is likely to exist in seagrass
ecosystems. This feedback indicated that seagrasses are able to trap and
stabilize suspended sediments, which in turn improves water clarity and seagrass
growth conditions. Furthermore, our analyses demonstrated that effects of
eutrophication on light conditions, as indicated by surface water total nitrogen,
were on average at least as important as sediment conditions. This suggests that
in general, eutrophication might be the most important factor controlling
seagrasses in sheltered estuaries, while the seagrass-sediment-light feedback is
a dominant mechanism in more exposed areas. Our study demonstrates the
potentials of SEM to identify and quantify positive feedbacks mechanisms for
ecosystems and other complex systems.

36

Gambar
. Diagram of the three stable and significant structural
equation models.
A) Model 1 provided the best fit to the data. It includes a positive
feedback loop between sediment grain size (D10), light attenuation
and seagrass density. B) The second best model (2) describes the
relation between seagrass and sediment grain size in the opposite
direction compared to model 1. C) The third, lowest-ranking model
(3) was the simplest of the three model and did not include a direct
relation between seagrass and sediment grain size. Note that light
attenuation was reciprocally transformed and that the effect of
nitrogen is therefore negative, while the effect of sediment D10 is
positive (higher value for light attenuation = increased water
clarity). Values above the arrow lines depict the standardized
regression weights.

4..

37

SEM-specific software
Scholars consider it good practice to report which software package
and version was used for SEM analysis because they have different
capabilities and may use slightly different methods to perform similarlynamed techniques (Kline, 2011).

Open source software

o
R has numerous contributed packages (e.g. OpenMx); see a
list of packages related to psychometric methodologies, including SEM.
o
nyx
o
WebSEM is an online software for SEM analysis

Commercial packages
o
AMOS in SPSS
o
EQS [1]
o
LISREL
o
MPlus
o
PLS-GUI
o
SAS (software) procedures
o
SmartPLS - Next Generation Path Modeling
o
Stata sem
o
WarpPLS
DAFTAR PUSTAKA
1. Kaplan, D. 2007. Structural Equation Modeling. Sage. pp. 79-88.
2. Kline, R. 2011a. Principles and Practice of Structural Equation
Modeling (Third ed.). Guilford.
3. Kline, R. 2011b. Principles and Practice of Structural Equation
Modeling (Third ed.). Guilford. p. 230-294.
4. Kline, R. 2011c. Principles and Practice of Structural Equation
Modeling (Third ed.). Guilford. p. 265-294.
5. Hancock, G. R. 2003. Fortune cookies, measurement error, and experimental
design. Journal of Modern Applied Statistical Methods, 2(2): 293-305.
6. Thorndike, R. 2007. Intelligence Tests". In Salkind, Neil. Encyclopedia of
Measurement and Statistics. Sage. pp. 477480. ISBN 9781412952644.
7. MacCallum, R. dan J.Austin. 2000. "Applications of Structural
Equation Modeling in Psychological Research". Annual review of
psychology 51: 201226.
8. Jansen, M., R.Scherer dan U.Schroeders. 2015. Students' self-concept and
self-efficacy in the sciences: Differential relations to antecedents and
educational outcomes". Contemporary Educational Psychology 41: 1324.
doi:10.1016/j.cedpsych.2014.11.002.
9. Gillespie, D. dan B.Perron. 2007. Structural Equation Modeling". In
Boslaugh, Sarah. Encyclopedia of Epidemiology. Sage. pp. 10051009.
ISBN 9781412953948.
10. Markus, K. 2007. Structural Equation Modeling". In Rogelberg, Steve.
Encyclopedia of Industrial and Organizational Psychology. Sage. pp. 774
777. ISBN 9781412952651.

38

11. Shelley, M. 2007. Structural Equation Modeling". In English, Fenwick.

Encyclopedia of Educational Leadership and Administration. Sage.
ISBN 9781412939584.
12. MacCallum, R. dan J.Austin. 2000. "Applications of Structural
Equation Modeling in Psychological Research". Annual review of
psychology 51: 201226.
13. Kline, R. 2011. Principles and Practice of Structural Equation
Modeling (Third ed.). Guilford. p. 205.
14. Kline, R. 2011. Principles and Practice of Structural Equation
Modeling (Third ed.). Guilford. p. 206.
15. Hu, L. dan P. Bentler. 1999. "Cutoff criteria for fit indexes in
covariance structure analysis: Conventional criteria versus new
alternatives". Structural Equation Modeling 6 (1): 155.
16. Browne, M. W. dan R. Cudeck. 1993. Alternative ways of assessing
model fit". In Bollen, K. A.; Long, J. S. Testing structural equation
models. Newbury Park, CA: Sage.
17. Hu,L. dan P.Bentler 1999. "Cutoff criteria for fit indexes in
covariance structure analysis: Conventional criteria versus new
alternatives". Structural Equation Modeling 6 (1): 155.
18. Kline, R. 2011. Principles and Practice of Structural Equation
Modeling (Third ed.). Guilford. p. 201.
19. Loehlin, J. C. 2004. Latent Variable Models: An Introduction to Factor,
Path, and Structural Equation Analysis. Psychology Press.
20. MacCallum, R. 1986. Specification searches in covariance structure
modeling. Psychological Bulletin, 100, 107-120.
21. Quintana, S.M. dan S.E. Maxwell. 1999. Implications of recent
developments in structural equation modeling for counseling
psychology. The Counseling Psychologist. 1999;27:485527.
22. Westland, J. C. 2010. Lower bounds on sample size in structural equation
modeling". Electron. Comm. Res. Appl. 9 (6): 476487.
23. Chou, C. P. dan P.Bentler. 1995. Estimates and tests in structural equation
modeling". In Hoyle, Rick. Structural equation modeling: Concepts,
issues, and applications. Thousand Oaks, CA: Sage. pp. 3755.
24. Bentler, P. dan C.P.Chou. 1987. Practical issues in structural equation
modeling". Sociological Methods and Research 16: 78117.
25. MacCallum, R. C., M. Browne, dan H.Sugawara. 1996. Power analysis
and determination of sample size for covariance structural modeling".
Psychological Methods 1 (2): 130149. doi:10.1037/1082-989X.1.2.130.
Retrieved 24 January 2015.
26. Judea.P. 2000. Causality: Models, Reasoning, and Inference. Cambridge
University Press. ISBN 0-521-77362-8.
27. Kline, R. 2011. Principles and Practice of Structural Equation
Modeling (Third ed.). Guilford. p. 79-88.
Bagozzi, R. dan Y.Yi. 2012. Specification, evaluation, and interpretation of
structural equation models". Journal of the Academy of Marketing
Science, 40 (1), 834. doi:10.1007/s11747-011-0278-x

39

Bartholomew, D.J. dan M.Knott. 1999. Latent Variable Models and Factor
Analysis Kendall's Library of Statistics, vol. 7. Arnold publishers,
ISBN 0-340-69243-X
Bentler, P.M. dan D.G.Bonett. 1980. Significance tests and goodness of fit
in the analysis of covariance structures. Psychological Bulletin,
88, 588-606.
Bollen, K.A. 1989. Structural Equations with Latent Variables. Wiley, ISBN
0-471-01171-1
Byrne, B.M. 2001. Structural Equation Modeling with AMOS - Basic
Concepts, Applications, and Programming.LEA, ISBN 0-80584104-0
Goldberger,A.S. 1972. Structural equation models in the social sciences.
Econometrica 40, 979- 1001.
Haavelmo, T.
1943. The statistical implications of a system of
simultaneous equations," Econometrica 11:12. Reprinted in D.F.
Hendry and M.S. Morgan (Eds.), The Foundations of Econometric
Analysis, Cambridge University Press, 477490, 1995.
Hair,J.F., G.T.M.Hult, C.M.Ringle dan M.Sarstedt. 2014. A Primer on
Partial Least Squares Structural Equation Modeling (PLS-SEM).
Thousand Oaks, CA: Sage. ISBN 9781452217444.
Hoyle, R.H.(ed). 1995. Structural Equation Modeling: Concepts, Issues,
and Applications. SAGE, ISBN 0-8039-5318-6
Hu, L. dan P. Bentler. 1999. "Cutoff criteria for fit indexes in covariance
structure analysis: Conventional criteria versus new alternatives".
Structural Equation Modeling 6 (1): 155.
Jreskog, K. dan F.Yang. 1996. Non-linear structural equation models: The
Kenny-Judd model with interaction effects". In G. Marcoulides and
R. Schumacker, (eds.), Advanced structural equation modeling:
Concepts, issues, and applications. Thousand Oaks, CA: Sage
Publications.
Kline, R. 2011. Principles and Practice of Structural Equation Modeling
(Third ed.). Guilford.
Kaplan, D. 2000.
Structural Equation Modeling: Foundations and
Extensions. SAGE, Advanced Quantitative Techniques in the
Social Sciences series, vol. 10, ISBN 0-7619-1407-2
Kaplan, D. 2007. Structural Equation Modeling. Sage. pp. 10891093..
MacCallum, R. dan J.Austin. 2000. "Applications of Structural Equation
Modeling in Psychological Research". Annual review of
psychology 51: 201226.
Brown, E.C., J.D.Hawkins, M.W.Arthur, R.D.Abbott dan M.L.V.Horn. 2008.
Multilevel analysis of a measure of community prevention collaboration,
American Journal of Community Psychology, 41, pp. 11526.
Little, T.D. 2013. Longitudinal Structural Equation Modeling, New York, NY,
Guilford Press.
Little, T.D., N.A.Card, D.W.Sledgers dan E.C.Ledford++. 2007. Representing
contextual effects in multiple-group MACS models. Dalam: T.D. Little, J.
A.Bovaird dan N.A.Card (eds). Modeling Contextual Effects in
Longitudinal Studies, Mahwah, NJ, Lawrence Erlbaum Associates.

40

Guo, B., B.E.Perron dan D.F.Gillespie. 2009. A systematic review of structural

equation modeling in social work research. British Journal of Social Work,
39(8): 15561574.
Rigdon, E.E. 2001. Structural equation modeling. Dalam: G. A. Marcoulides (ed.),
Modern Methods for Business Research, Mahwah, NJ, Lawrence
Erlbaum.
Okech,D., J.Kim dan T.D.Little. 2013. Recent Developments in Structural
Equation Modelling Research in Social Work Journals. British Journal of
Social Work (2013) 120. doi:10.1093/bjsw/bct154.