Summary of statistical techniques in empirical analysis
Once the data collection is over, the selection of appropriate statistical techniques logical next step to analyse the data (Tabachnick and Fidell, 2007) and to study the relationships between the variables (Table 5.5). The domains of techniques may be classified as: univariate, bivariate and multivariate statistics.
Univariate statistics Analyses in which there is a single variable (primary for descriptive statistics)
Bivariate statistics Analysis of two variables simultaneously (study the relationship between the variables)
statistics Analyses of multiple variables simultaneously (study the relationships of multiple dependent
and independent variables)
variables (DV) Variables thought to be influenced by other variables to behave in a certain way.
variables (IV) Variables that cause a reaction in the DV, or it may explain why the DV fluctuates. It is thus
assumed that changes in the IV will usually precede any change in the DV.
Table 5.5 Domains of statistical techniques and variables (Babbie, 2004; Tabachnick and Fidell, 2007)
Multivariate statistical techniques are recommended for analysing complex data (Flynn et al., 1990) for the research in production management. Forza (2002) and Scudder and Hill (1998) enlisted the several multivariate statistical techniques which are useful in empirical data analysis (Table 5.6).
Multivariate statistics techniques Purpose
Multiple regression Predict the changes in the Dependent Variables (DV) in response to changes in
the several Independent Variables (IV).
Multiple discriminant analysis To understand group differences and predict the likelihood that an entity (individual or object) will belong to a particular class or group based on several
Multivariate analysis of variance/covariance
(MANOVA/MANCOVA) To simultaneously explore the relationship between several IVs (usually categorical data) and two or more DVs.
Cluster analysis To classify a sample of entities (individuals or objects) into a smaller number of
mutually exclusive subgroups based on the similarities among the entities
Factor analysis To analyse interrelationships amongst a large number of variables and to explain these variables in terms of their common underlying dimensions (factors).
Structural equation modelling To simultaneously test the measurement model (which specifies one or more
indicator to measure each variable) and the structural model (the model which relates causal inference between multiple IVs and DVs).
Table 5.6 Main multivariate statistics techniques (Forza, 2002, p186)
Figure 5.4 The choosing of the techniques based on Walker and Maddan (2012, p456)
On the basis of the research questions and the data collected, this current study follows Walker and Maddan (2012) to adopt factor analysis to identify underlying factors of the questionnaires measurement instruments and applies the structural equation modelling to test the hypothesised causal relations (Figure 5.5).
A methodical approach of the research method
On the basis of the approaches proposed by Forza (2002), Oppenheim (1992), and Flynn et al. (1990), a methodical approach was used for guiding the research process and data collection (Figure 5.5).
Figure 5.5 A methodical approach following Forza (2002, p157), Oppenheim (1992, p7), and Flynn et al. (1990, p254)
Link to the theoretical level
This is the first step. Here, the aims and objectives of the study were decided which were presented in Chapter 1. The theoretical framework was established after a critical analysis of the literature which was elaborated in Chapter 2 and 3. The development and articulation of the instruments of data collection, i.e., the theoretical framework and research hypotheses were discussed in Chapter 4.
Designing the research
The aim of this step is to justify the research process design, strategy and data collection method. The reasons behind selecting the empirical deign and survey strategy were presented in the Chapter 5 (Section 5.1.1 to 5.1.3 of this chapter).
Developing instruments for data collection
The explanation about the adoption and development of the measurement instruments in the questionnaire is the objective of this step. Since the constructs in the production management are complex and have multiple facets which generally cannot be measured directly (Forza, 2002), multi-items instruments are recommended for accurate and complete measurement (Hensley, 1999). This current research followed the processes suggested by Oppenheim (1992), Malhotra and Grover (1998), Forza (2002), and Saunders et al. (2007) to develop the measurement instruments (Table 5.7). The details of these processes are elaborated in Chapter 6.
Define the questions (Wording) Specify research domain based on existing literature to ensure content validity (the adequacy with which a measure or scale has sampled from intended domain of content).
Formulate the questions with clear interpretations. They can be adopted, adapted or newly developed.
Choose from open-ended (allowing respondents to answer in any way they choose) or
closed (limiting respondents to a choice among alternatives given by the researcher) questions.
Decide the scales (Scalding) Decide measurement scales to be used to measure the answers. The four basic types of scale are nominal (categorically discrete data, i.e., names), ordinal (quantities that have a natural ordering, i.e., Likert-scale), interval (similar to ordinal but each value are equally split, i.e., temperature), and ratio (similar to interval with a natural 0 point, i.e.,
fixed sum scale).
Identify the respondents (respondent identification) Identify the population (the entire group of people, firms, plants or things that the researcher wishes to investigate), sampling frame (a listing of all the elements in the population), and the sample (a subset of the population, it comprises some members selected from the population) of the research.
Select the appropriate informants to collect data. Probability sampling (or representative sampling) is the most commonly used method in survey strategy. It comprises four stages: identify a suitable sampling frame; decide a suitable sample size; select the sample; and check for the representative of the population.
Put together and test questions
(Rules of questionnaire design) Construct the questionnaire, including the layout, order and flow of the questions, and translate the questions into other languages (if needed).
Pilot test the questionnaire.
Table 5.7 Processes to develop questionnaires (Flynn et al., 1990; Oppenheim, 1992; Malhotra and Grover, 1998; Forza, 2002; Pallant, 2007; Saunders et al., 2007; Karlsson, 2009)
Collecting and screening of data
This present research followed the processes evolved by Hair et al. (2010) and Tabachnick and Fidell (2007) to screen the data, obtain descriptive statistics and handle the non- response bias using SPSS (Table 5.8). The Chapter 6 presents these processes in detail. Factor analysis was adopted to manage the data, analyse the interrelationships among the variables (to uncover the underlying factors) and evaluate these relationships (to ensure construct validity, the extent to which the factors were correctly correlated to measure the reality). The steps involved in performing the factor analysis are discussed in Section 5.2. The results and measurement quality assessment are presented in Chapter 6.