1

Int J Adv Manuf Technol (2015)

Int J Adv Manuf Technol (2015)

DOI **.****/s*****-***-****-*
ORIGINAL ARTICLE

Statistical testing of structural model and its critical factors through a

simulated success latent fourth variable for industrial projects
management.
H. Garca MC1& Dr. A. Valles1& Dr. J. Snchez
G. Dominguez
Received: 201 / Accepted: 201 / Published online: 201
# Springer-Verlag London 201
a
,
Me
xico.
e-mail:hugarci_20@yo.m

Abstract: The literature and viewed items on this subject

show no statistical data examination regarding the
relationship or approach between critical factors (CF) and
their association or significance for project evaluation and
corroborate for successful realization.
Executives, project managers and staff members have no
statistical data on the matter and importance regarding CF
and the influence for the possible outcomes and scenarios
these factors and criteria yield neither a basic structural
model on project management and achievement rationale to
conduct their time, process standards, efforts and resources.
H.GarcMC
oPfesr
pdartme
n
t
,
Ciuda
d

at

nduIstrial

ngier

Juare
z
Instiu

hPD.AVales
Resarch
Ciuda
t
,
Ci
uda
Chiua
a
,
Me
xico.
hPD.JSnezc
Resarch
pdartme
n
t
,
Ciuda
d

opfesr

at
Juarez

nduIstrial

engir

Instiue

Juarez

Instiue

profes

depnartm

of

of

Technolgy,

at

Technolgy

Cd.

Juarez,

Industrial

engir

Juare

e
z

Instiu

o
f

e
Technolg

y
,
Av
e
.
Tecnol
gico
Chiua

o
f
Technolg

No.

409,

Cd.

Juarez,

y
,
Av
e
.

Tecnol J Adv Manuf Technol (2015)

Int
gico
No.
409,
Chiua
a
,
Me
xico.

Cd.

Sg.Noarie
oPfesrnatduIilgmp,Autonomous University of Ciudad Juarez, Ave. del
Charro 450 Norte, 32310 Cd. Juarez, Chihuahua,
Mexico.

Int J Adv Manuf Technol (2015)

Juarez,

1-Defining
the
Research
Question
6Summarizat
ion

2-Location
of Research
Studies .

Methodolo
gy Metaanalysis

For this work were analyzed the basic formats of

scientific articles under definite and established
methods, to develop a scientific study and strategies to
determine an evaluated a structural model, with the
goal analysis methodology for the management on
industrial projects, identifying: sequence patterns,
articles, classification factors, calculations frequency
factors, and statistical relationships between them
within the proposed path diagram structural model,
and simulated regarding success.

5- Quality
Assessment
of Studies

3- Criteria
for item
selection

4Identifying
Studies

Keywords Critical success factors (CSF), evaluation

criteria, meta-analysis, structural modeling, project
management,
exploratory
factor
analysis,
confirmatory factor analysis, latent variables.
1 Introduction & bibliographic revew
Itispner&ogmdf hjctaerizn,smypovdutakecrniqos(American National Standards Institute /ProjecntMmagIsiu)[1],NA(204.Twpctenardlohgsipce,wmtofndriaspctlon;CSF(riues)actdKRAylr.ThmosefivuCSFcnltgwxrkpeid,oyhfcatsrnigodmpevatnbusid(Tryh).cevoalztnfuKdwhoscelmptir(vnbhkgao),fsdetmirplcnaovsemht.Trfipnayzletuohgrsmdifncktalorswhemipngatofrjecshpiumn-alytodgbsicerpvhwndftrelaiopbsguc.

Figure 2.1 Methodologies [2].

2 Methods
Meta-analysis (MA) takes a large quantity of
quantitative extracted results on individual investigations,
with the purpose of outcome integration and improve
understanding, this systematic revision tries to reach
conclusions from a variety of studies on the same subject; by
analyzing procedures and standardized techniques. This
statistical method is fundamental to delimitate the success
criteria for industrial projects.
There is no single method for MA preparation, particular
studies excludes or adds some stages, however, definite
steps are listed next and make the algorithm followed in this
study. [5].
The methodology consists of six stages which are:

2.1 Definition of structural equation modeling (SEM)

This technique combines factor analysis with linear
regression to test the observed goodness of fit of the
collected data to a hypothesized model and expressed by the
trails diagram. As a result, SEM provides the values within
each relationship, and most importantly, a statistic that
expresses the degree to which the data fit the proposed

model,
confirming
its validity.
Int J Adv
Manuf Technol
(2015) Among the strengths of SEM
is the ability to construct latent variables, that are not
directly measured, but are estimated in the model from a
number of variables that covariance with each other. This
allows to explicitly capture the reliability of the model.
Factor analysis, path analysis and linear regression represent
special cases of structural equation model. [3]. It is a statistic
meant for testing and estimating causal relations from
statistical data and qualitative causal assumptions art.
[4].This definition has been articulated by the geneticist
Sewall. [6].

Int J Adv Manuf Technol (2015)

The structural equation models were born from the need to

provide more flexibility in the regression models. They are
less restrictive than the regression models allowing for the
fact to include measurement errors in both criterion
variables (dependent) as the predictor (independent)
variables. The Structural Equation Modeling methodology is
a very young area for developing statistics compared to
regression models and factor analysis.
SEM has a confirmatory nature more than an exploratory; it
is part of a relevant theoretically hypothesis in the context of
interest. A major strength for this approach is the ability to
develop constructs that estimate latent variables based
around measurable variables.
2.2 Applying the statistical analysis methodology SEM
These 16 critical factors show a correlation covariance table
between pairs of factors for each of the three criteria to
evaluate: finish on time the industrial project, fulfill the
budget of the industrial project and fulfill the quality of the
industrial project, with a total of 256 correlations for each
evaluation criteria, where the most significant correlations
with P-value of P <0.01 level are indicated, as shown in
Tables 2.1, 2.2 and 2.3 giving a total of 11 critical factors
considered for model building. All these correlation values
and covariance were calculated with SPSS 22 statistical
software and they are shown below on table 2.4.

Table 2.2 Correlations for the latent variable 'fulfall the

quality of the industrial project'

Table 2.1 Correlations for the latent variable 'finish the

project on time'

Table 2.3 Correlations for the latent variable fulall the

budget of the industrial project

Int J Adv Manuf Technol (2015)

through meta-analysis,
by frequency
methodology
Int J Advclassified
Manuf Technol
(2015)
4
and empirical results obtained from surveys conducted in
companies of the region, which are evaluated by statistical
tools such as exploratory factor analysis and confirmatory
factor analysis to determine the structural equations where
the success criteria are also related. In the figure 2.1 we can
see the first structural model proposed.

Table 2.4 Critical success factors according to the bivariate

correlation value

Figure 2.2 Path Diagram Proposed Structural Models

2.3 Development of a good theoretical model

This are the most important critical factors proposed
resulting from the theoretical analysis of the factors found

2.4 Factor analysis

Factor analysis divides the variance of each indicator
(derived from the correlation matrix and sample covariance)

inInttwo
parts:
(1)Technol
"common
variance", the variance explained
J Adv
Manuf
(2015)
by the latent (s) variable (s), which it is estimated on the
basis of shared variance with other indicators in the analysis;
and (2) "unique variance" which is a combination of a
specific reliable variance for the indicator and a random
error variance.
There are two main types of model-based analysis of
common factors: Exploratory factor analysis (EFA) and
confirmatory factor analysis (CFA). Both tests are designed
to reproduce the observable relationships between a set of
indicators with a small set of latent variables. However, EFA
and CFA differ primarily in the number and nature of a
priori specifications and restrictions made in the
measurement model of the latent variable. EFA (initially) is
data-driven so that no specifications made in relation to the
number of common factors or the pattern of relationships
between common factors approach. Rather, the researcher
uses the EFA as an exploratory or descriptive technique to
determine the appropriate number of common factors, and to
check which measurement variables are reasonable
indicators of various latent dimensions.
In CFA, the researcher specifies in advance the number of
factors and pattern of load indicators factors; other
parameters, such as those bearing independence or
covariance factors and unique variances and indicators

For this part Matrix

variance
is explained
Int J Adv
Manuf Technol
(2015) by the principal
5
components method where the left section of the table
displays the variance explained by the shown initial solution.
Only the major factors in the initial solution are greater than
1. Overall eigenvalues, representing almost 65% of the
variance for the original variables. This suggests that the
three latent influences are associated with the use of the
service, but there is room for a lot of unexplained variation.
The second section of the table shows the variance explained
by the factors taken before rotation. The cumulative
variability explained by these three factors in the extracted
solution is about 55%, a difference of 10% of the initial
solution. Thus, about 10% of the variation explained by the
initial solution is lost due to single latent factors to the
original variables and variability that simply cannot be
explained by the model of factors.
The rightmost section of this table shows the variance
explained by the factors extracted after rotation.
Table 2.5 Total explained variance matrix

2.5 Exploratory Factor Analysis

The exploratory factor analysis (EFA) is a statistical
approach to determine the correlation between variables in a
dataset. This type of factor analysis provides a structure (a
grouping of variables based on the strong correlations). In
general, an EFA prepares the variables to be used to clean
structural equation models. An EFA should always be
conducted for new datasets. The advantage of an EFA on a
CFA (confirmatory factor analysis) is that no theory is
applied a priori to which elements belong to which
constructs. This means that the EFA is able to detect
problematic variables much more easily than the CFA. In
this section we will develop EFA for this search.

2.5.1 Explained variance matrix by the method of principal

components

2.5.2 KMO and Barletts test

This
shows
two tests
Int J table
Adv Manuf
Technol
(2015)which indicate the suitability of
the data for the detection of the structure. The sampling
measure of adequacy Kaiser-Meyer-Olkin is a statistic that
indicates the proportion of variance in the variables that may
be caused by underlying factors. The high values (about 1.0)
generally indicate that an analysis of factors may be helpful
with your data. If the value is less than 0.50, the results of
the factor analysis are unlikely to be useful.

Int J Adv Manuf Technol (2015)

Table 2.6 Initial KMO

Table 2.7 Final KMO

2.5.3 Pattern Matrix

The elements of this matrix are called weights or loads.
Indicate the weight between each observed variable and
each factor. It should be interpreted as standardized slopes
(beta) in a multiple regression analysis.
This section describes the configuration tables thrown
through analysis shows, as well as the adjustments made
through the exploratory factor analysis.

Table 2.8 Initial pattern matrix

Table 2.9 Final pattern matrix

Int J Adv Manuf Technol (2015)

"Confirmatory Factor
Analysis"
(CFA)(2015)
is a type of structural
Int J Adv
Manuf Technol
7
equation modeling that deals specifically with measurement
models, that is, the relationship between observable
measures or 'indicators' and latent variables or "factors". The
measurement objective of the models for latent variables sets
the number and the nature of the factors considered for the
variance and covariance by a set of indicators. One factor is
an unobservable variable that has greater influence than an
observable measurement and is considered for correlations
across these observable measurements.
The load factors dropped by the configuration matrix in the
exploratory factor analysis, which are used to construct the
first model using AMOS, shown in figure 2.2 for the initial
proposed model and figure 2.3 for the final model with its
covariance on error variables e31 and e33.
2.6.1 Goodness Fit indexes
P-value: A p-value of 0.05 is common used in sociological
research. The obtained error can also be addressed by
increasing the size of the sample, reducing the possibility
that the obtained data is coincidentally rare.
Comparative Fit Index (CFI): A rule of thumb for CFI and
other comparative fit index for a value greater than 0.90 can
reasonably indicate a good fit in the researcher's model [8].
Root Mean Square Error of Approximation (RMSEA): The
RMSEA is an index of "badness of fit" in which a value of 0
indicates better fit and higher values indicate poorer
adjustment.
RMSEA0.06 indicates close fit proximity. Values between
0.06 and 0.08 suggest a reasonable fit error. And RMSEA1
suggests a poor fit. [7].
2.6.2 Modification Indices

To interpret the factor solution the Matrix configuration

must be interpreted. Performance mode is similar to the
matrix in the orthogonal rotated factors case. The oblique
case (the most important of all matrices obtained); by setting
Matrix and Matrix of correlation between the factors. Thus
necessarily to be included in the results analysis.

2.6 Confirmatory Factor Analysis

Each parameter has a rate above certain beginning

modification. Not displayed modification indexes indicate
that they dont exceeds the specified threshold.

2.6.3 Initial Results.

AtIntthis
point
theTechnol
initial(2015)
proposed model with its regression
J Adv
Manuf
loads and the goodness fit indexes are shown in the figure
2.2.

Int J Adv Manuf Technol (2015)

Table 2.16 Initial modification indices

2.6.4 Partial Results

Figure 2.3 Initial model and its standardized estimations

Table 2.10 Initial P-value

For the partial model with its regression loads and the
goodness fit indexes are shown next. For this partial model
the variables error for e31 and e33 were covariate as the
modification indexes showed the highest relation between
them (table 2.16). This action generate a better modeling fit
compared with the first model, and confirmed by the tables
for goodness fit indices.

Table 2.12 Initial CFI

Figure 2.4 Partial model and its standardized estimations

Table 2.11 Partial P-value

Table 2.14 Initial RMSEA

Int J Adv Manuf Technol (2015)

In table 2.17 we
the
high (2015)
relation among 9error
Int Jcan
Adv see
Manuf
Technol
variables e26 and e31, however to covariate them is
impossible because they are in different indicators set.

2.6.5 Final Results

Figure 2.4 shows the construction for a regression loads
model with fore latent factors, adding an additional
simulated latent variable (success), on table 2.23 & 2.24
indicated the regression loads for each relation with in the
model. Also the three goodness of fit indicators show an
improved proportion compared to the previous models.

Table 2.13 Partial CFI

Figure 2.4 Regreccions Loads between Critical Factors and

the Proyect Succes Criteria.

Table 2.18 Final P-value

Table 2.15 Partial RMSEA

Table 2.17 Partial modification indices

Table 2.19 Final CFI

Table 2.20 Final RMSEA

Int J Adv Manuf Technol (2015)

Int J Adv Manuf Technol (2015)

10

Standardized regression weights:

Table 2.22 Final modification indices

The correlation estimates and the standardized regression

loads is re-calculated, as an example when quality rises 1 for
standard deviation, Q.COMUNICATION is adjusted at
0.302 standard deviation on table 2.24.
Table 2.23

If the analysis is repeated the discrepancy between e7 and

e10 is reduced at least 5.013 for the covariance treatment as
a free parameter.
The p Value obtains 0.148 for partial results because this is
an initial result after the covariance on e31 and e33 however
the CFI and RMSEA improved, and for the simulated latent
variable run this p value decreases to 0.032.
2.7 Hypothetical regression structural equation modeling
The traditional approach to integrating multiple regression
analysis and factor analysis involves factoring a set of
indicators of one or more predictors and outcomes,
generating factor scores (which, as noted, are indeterminate)
or creating unit-weighted composited of the highest-loading
indicators, then using these variables as predictors or
outcomes. SEM allows for these two components of the
analytic strategy to be done simultaneously; that is, the
relations between indicators and latent variables and the
relations between variables are evaluated in a single model
[7]. Here we develop the equation model for this research.
F1 = 1 Latent variable Finish the Project on time
F2 = 2 Latent variable Fulfill the quality of the Industrial
Project.
F3 = 3 Latent variable Fulfill the budget of the Industrial
Project
x = Coefficient between an observed variable or between a
latent variable and an observed variable.
bx = Coefficient between latent dependent variables.
Regression equations:

Structural model:

Table 2.24

Int J Adv Manuf Technol (2015)

It is recommended
methodology
and replicate
Int J to
Advapply
Manufthis
Technol
(2015)
11
research in other sectors and industries to test the generality
of the model.
5. References

3. Conclusions

1.

ANSI/PMI 99-001-2004 (2004), N. A. Fundamentos de la

Direccin de Proyectos. USA: Project Management Institute
Tercera ed.

2.

2. Coln (2007), E.N. Tesis: Factores Organizacionales que

Impactan en la Aplicacin Exitosa de Clulas de
Manufactura. Instituto Tecnolgico de Cd. Jurez.

3.

Iacobucci, D. (2010), Structural equation modeling: fit

indices, sample size, and advanced topics. Journal of
Consumer Psychology, 20, 2010, 9098.

This contribution to the theory of project management is

novel because by meta-analysis of a large number of
indicators tentatively filtered the most important final
indicators are obtained and statistically defined by SEM
methodology and its relations with respect to the latent
variables.

4.

Iacobucci(2009), D. Everything you always wanted to know

about SEM (structural equation modeling) but were afraid to
ask. Journal of Consumer Psychology, 19, 2009, 673680.

5.

Valles, A. (2008) Factores de Mayor Influencia en la

Efectividad de Tcnicas de la Manufactura Esbelta. Tesis
Doctoral, Instituto Tecnologico de ciudad Juarez.

4. Recommendations

6.

Wright, S. Anecdotal (1999), Historical and Critical

Commentaries on Genetics. Edited 4 James F. Crow and
William F. Dove.

7.

Hoyle, R. (2012) Structural Equation Modeling Handbook.

8.

Hu and Bentler (2006), Structural equation and log-linear

modeling: a comparison of methods in the analysis of a
study on caregivers' health.

As can be seen, there are tests for goodness of fit with the
confirmatory analysis of the greatest loads of indicators
(endogenous variables), showing which of them should take
more care, if we want to modify the subsequent results
values of the latent variables, will be improved conditions
and consequently endogenous variables decrease the size of
the error in each case.

Indications of interest were observed in the survey of factors

and most important criteria, like the exogenous variable
customer satisfaction. It is recommended to continue this
research by adding the exogenous latent variable customer
satisfaction and their respective survey and change the
model trajectories.