You are on page 1of 6

CHAPTER 3

RESEARCH DESIGN
3.1 INTRODUCTION

The second chapter gave a comprehensive review of earlier research conducted in the area of
small and medium enterprises (SMEs).The review of literature clearly established the need for
identification of key success factors and barrier factors for successful implementation of lean
manufacturing in small and medium enterprises (SMEs). Rarely do we find that any attempts
have been made to study the relationship between success factors, barrier factors and its
contributing variables. There is not much literature on establishing a model linking all these
factors using the structural equation modelling approach.

From fourth to ninth chapter, the theoretical foundation of the conceptual framework for the
research work has been outlined. In this chapter, the research design adopted for establishing
the model, which links the success and barrier factors and the contributing variables by the
structural equation modeling approach, is presented.

3.2 SAMPLE DESIGN

The sample for the study is the small and medium scale organizations in Kerala a state of India.
Random sampling and convenience sampling have been used for selecting the samples of the
study. The sample size has to be appropriate while using the structural equation modeling
approach. It is generally understood that the SEM requires larger sample sizes (Kline 2005).The
estimation of more statistical effects may require more complex models and in order to have
rationally stable results, larger sample size is necessary. The sample size requirements also
affect the type of estimation used in the analysis. In SEM, there is more than one type of
estimation and because of the assumptions made about the data some of these may require large
sample sizes .Kline (2005) pointed out that any type of SEM analysis with sample size less
than 100 cases may be untenable unless a very simple model is evaluated. If size of the sample
is less than 100, then it would be considered small. A sample size between 100 and 200 is
considered minimum. Sample size is not infinite because things such as complexity of the
model must be considered. A sample size more than 200 cases would be considered as large

Breckler(1990) has given another empirical guide line about sample size. He has surveyed 72
published journals in personality and social psychology in which some type of SEM was
conducted. The sample size across these was 198, according to the guideline given by Kline

25
(2005) which is approximately medium. Bollen (2000) who surveyed the application of SEM
published in 16 different research journals from 1993 to 1977 observed that 20 percentage of
studies used the sample size less than 100 cases.

McQuitty (2004) has suggested that, when SEM is used it is important to determine the
minimum size required prior to the data collection in order to achieve the desired level of
statistical results with a given model. Schuman et al (2006) have pointed out that normality of
data and estimation method used by researches affected by the sample size, although the
generally agreed value of sample size is 10 members for every free parameter estimated.
However Sivo et al (2006) have proposed a critical sample size 200 for SEM. Schuman et al
(2006) have pointed out that normality of data and estimation method used by researchers
affected by the sample size, although the generally agreed value of sample size is 10 members
for every free parameter estimated. However in this research same criteria as put forward by
Schuman et al (2006) has been used for the sample selection.

Research required a two-stage questionnaire, the first questionnaire is developed to identify the
success factors and barrier factors for the implementation of lean manufacturing in small and
medium enterprises and second questionnaire is suitable for modeling the relationship between
each variable in success factors and barrier factors. Both the actual questionnaires
administrated to the small and medium enterprises are given in the appendix of this report.

3.3 PLACE OF STUDY

This study was conducted in the geographical area of Thrissur in the state of Kerala in India.
In the state of Kerala Thrissur is considered one of the major industrial hub after the city of
Kochi. The city’s industrial growth started in 1956 and accelerated after the 1980’s.TheMajor
type of industries located in Thrissur include, plastic industry, die manufacturing industry,
forging industry and pen industry. More significantly, the city of Thrissur has the distinction
of being one of the active commercial centers of Kerala, with excellent rail, road, and airline
connectivity to all major metros of India and to many regional centers too. Another reason for
selecting Thrissur city as the place of study is that the investigator located here and familiar
with the area.

3.4 DATA COLLECTION PERIOD

The data collection for this study by using structured questionnaire through face to face
interaction with the small and medium scale organizations was completed over a period

26
spanning fourteen months in two stages between October 2014 to April 2015 and between 2015
October to April 2016.From the month of October 2014 to April 2015 key success and barrier
factors for implementing lean in SMEs from first structured questionnaire was identified .The
period between 2015October and April2016 was used to establish a model linking the success
and barrier factors and its contributing variables through the second structured questionnaire.

3.5 PILOT STUDY

After finalizing the number of items in the questionnaire, a pilot study was under taken to assess
the reliability and validity of the questionnaire. The pilot study of the questionnaire was
conducted by administering the questionnaire to 30 organizations. For the reliability
assessment of the questionnaire, the internal consistency of the data collected from the pilot
study was subjected to the reliability test using Cronbach Alpha. Cronbach Alpha is the most
prominent reliability coefficient. It measures the reliability of a set of indicators. Its value
ranges from zero to one. Confirming the internal consistency the resultant reliability coefficient
of the variable chosen was found to be more than 0.6 (Sakakibara et.al, 2009).The analysis
during the pilot study also suggested that, the administrating of the questionnaire was practical
to bring out the necessary information and on average .Around 15 to 20 minutes are required
to complete the questionnaire.

3.6 DATA COLLECTION

Two sets of questionnaire were administrated and it was done sequentially in two phases as
indicated earlier. The sample of organizations have been obtained from Kerala, as no study
has been conducted to identify the key success factors for lean implementation SMEs here.
Small and medium size companies were randomly selected and face-to-face interviews were
carried out with the questionnaire. The respondents involved in this questionnaire were
managers, supervisors, and owner of companies. They are considered as suitable personnel
who are likely to be leaders in charge of lean implementation. However, here it has to be
mentioned that same respondent was not given both sets of questionnaire. Different sets of
questionnaires were given to different sets of respondents. This is because the second
questionnaire was developed after the analysis of the first questionnaire. In addition, it was
almost impractical to meet those respondents who answered the first set of questions to answer
the second set too, considering the four-month time lag between administrating the first and
second set of questionnaires.

27
3.7 STATISTICAL TOOL USED

A combination of univariate and multivariate statistical techniques are available for statistical
analysis. Out of this Confirmatory Factor Analysis, Exploratory Factor Analysis, Structural
Equation Modelling, and Specification Search were used for the data analysis of the study.
These techniques are described further.

3.7.1 EXPLORATORY DATA ANALYSIS

Exploratory factor analysis is a convenient method of simultaneously pointing out the group of
variable, which constitute independent response set based on their variability (Richard L.
Gorsuch 1997). In the present study principle component factor analysis (Varimax Rotated) is
applied to the success and barriers factors with 27 and 31 items. The purpose of confirmatory
factor analysis (CFA) is for labelling those items not loaded in any one of the dimension.

3.7.2 CONFIRMATORY FACTOR ANALYSIS

Confirmatory factor analysis (CFA) was used for the purpose of identification of key success
factors and barrier factors and their associated variables for the successful implementation of
lean manufacturing in SMEs. According to Clement and Israel (2005) CFA allow the
researcher to determine the number of factors and loading of measured variable on them
conform to what is expected on the basis of a pre-established theory. On the basis of prior
theory, indicator variables are selected and factor analysis is used to see if they load as predicted
on the expected number of factors. Amos SEM typically uses the CFA model to build the causal
relationship among the latent variable. The model is evaluated using goodness of fit measures
generated by the Structural Equation Model (SEM) packages. The main advantage of CFA
over the exploratory factor analysis (EFA) is that, CFA considers and eliminates the
measurement errors in relating variable to the dimension to that extent results are reliable.
Further, the number of variables attached to construct could be reduced by the researcher
through the confirmatory factor analysis. Unlike EFA, CFA results can be tested for validation,
which is indeed unique and same, is performed for the present study.

3.7.3 STRUCTURAL EQUATION MODELING

Structural Equation Modeling (SEM) is a modeling technique, which is widely used, in


behavioral research. Large number of endogenous and exogenous variables that are latent is
capable of handling in SEM. The variable that cannot be directly measured is called, as latent
variable therefore needs to specify through a combination of observed variable. SEM is

28
classified into two parts-one is measurement model and second is structural model. The relation
between the observed variable and latent variable are specified using measurement model. The
structural variable measures the impact of one variable on the other. Combination of factor
analysis, multiple regression, and canonical correlation are possible using SEM. In this
research, different models are developed using the AMOS (Analysis of Moment of Structures)
to test the hypothesis. SEM is primarily designed for path analysis, covariance structure
modeling, though it may be used to perform linear regression analysis, analysis of variance
(ANOVA) and analysis of covariance (ANCOVA)( Hair et al. 2006).AMOS provides an
intuitive graphical interface that allow the analyst to specify the model by drawing them. It has
also a built in bootstrapping routine and superior handling of missing data. It helps in
developing the models to test the hypothesis and determine the relationship among the observed
and latent variables .The SEM analysis relies on the number of fit indices to assess the fit of
measurement model which includes the i) Normed or relative Chi-square (CMIN/DF), (ii)
Comparative Fit Index (CFI), (iii) Goodness of fit index (GFI), and (iv) Root Mean Square
Error Approximation (RMSEA) (Hair et al. 2006). Secondly maximum likelihood estimation
is used for the full structural model measurement. The correlation links between the theoretical
variables are estimated and the coefficient value of the casual and the model fit is verified based
on the criteria of (Hair et al 2006)

3.7.4 SPECIFICATION SEARCH

Specification search was used in the research to develop the model.The modification of an
initial model to improve the fit has been termed specification search (Long, 1983).For
conducting the specification search the most common approach was to change the parameter
restriction one at a time in the initial model to examine model fit improvement. i.e. comparison
of full and restricted models. TETRAD, a program used in the specification of causal models
(Spirtes, Glymour, & Scheines, 1993), MallowsCp (Schumacker, 1994b), the Tabu search
(Marcoulides, Drezner, & Schumacker, 1998; Dreznoer, Marcoulides, & Salhi, 1999), and the
genetic algorithm procedure (Marcoulides&Drezner, 2001) are the most recent specification
search. Specification search procedure include the use of all possible sub set selection of
parameter built on Bentler& Chou’s(1990) and continue the specification search even when a
model results in non-significant fit function suggested by Mac Callum(1986).Amos 20
exploratory specification search procedure, guide the researcher on selecting the best fitting
model from all possible subset model with a set of fit function criteria.AMOS20 automated
exploratory search procedure provide the researchers with an easy implementation model

29
modification using several well-known fit function criteria. In the automated exploratory
search procedure, the top ten models are ranking by default given several fit function criteria.
The ranking of the model is similar to how AMOS ranks the model according to the other fit
measures i.e. ranking according to RMSEA, RFI, or TLI(Tucker Lewis Index) is same as
ranking according to chi-square/df (Arbuckle, 2003). However, the researcher must ultimately
choose which best model to retain

3.7.5 MODEL VALIDATION OF THE PREDICTIVE MODELS

All the results of the model have been validated using questionnaire .The same structured
questionnaire was used to collect the feedback of predictive models from three different
manufacturing SMEs (Annexure No 9) those SMEs are not part of the earlier studies. Using
SPSS the t test has been conducted. The t-test assesses whether the means of two groups are
statistically different from each other. The study shows that results are satisfactory at some
extents

30

You might also like