Chapter 3 Methods
3.3 Data Analysis
3.3.1 Factor Analysis
As the data set was relatively large and consisted of a number of different variables it was decided that the best way to handle it was to reduce or summarise the data using a smaller set of factors or components. Factor Analysis is a technique that looks for ‘clumps’ or groups among the correlations within a set of variables, something that is impossible to do with the naked eye (Pallant 2007: 179). The intention of the project, it will be recalled, was to answer the following research questions:
1) What are the predictive factors that formulate the attitudes of young Japanese people on whaling issues?
2) Of these predictive factors, which make the most significant contribution to the whaling attitudinal model of Japan’s youth?
Exploratory Factor Analysis (EFA) was chosen as the most appropriate technique as this method is specific to gathering information about the interrelationships among a set of variables. Confirmatory Factor Analysis was dismissed as this particular method is used to test specific hypotheses or theories concerning the structure underlying a set of variables (Pallant 2007: 179). In regards to sample size, the number of completed useable surveys was 529, well over the 300 that is deemed ‘comforting’ for factor analysis by Tabachnick and Fidell (2007: 613). Some authors
suggest that it is not the overall sample size that is of concern but the ratio of subjects to items. Thus, Nunnally (1978: 421) recommends a 10 to 1 ratio, with others suggesting a 5 to 1 ratio (see discussion in Chapter 1 of Tabachnick and Fidell [2007]). In the case of the factorial analyses performed in this project, the ratio between cases and items was 529 to 1, satisfying the ratio criteria more than adequately.
The answers to several selected survey questions were subjected to the Exploratory Factor Analysis (EFA) (Principal Component Analysis with Varimax rotation) in order to obtain a baseline understanding of any underlying concepts and interrelationships. The survey questions that were excluded from the EFA were the questions that provided finite answers (such as ‘Yes’, ‘No’ and ‘Not Sure’) and questions that sought information on personal experience of whale products (Questions 7-10). All of the remaining questions in the survey were subjected to an initial EFA. This analysis identified five constructs:
1. approval of whaling;
2. intangible motivations of whaling;
3. whale conservation;
4. maintaining the whaling industry;
5. and, acceptance of pro-whaling rhetoric.
A sixth construct, the approval of the consumption of whale meat by Japanese children, was created by calculating the mean scores of two questionnaire statements relating to this topic. Several of the questions were anti-whaling and needed to be reverse coded (Questions 4, 5, 6, 8, 9, 10, 12, and 13) to allow the analysis to focus specifically on positive attitudes to whaling.
(KMO) tests (Kaiser 1970, 1974) and Bartlett’s Test of Sphericity (Bartlett 1954) were both performed, in addition to inspections of correlation matrices (to examine the strength of the correlations among items). When using the KMO measure of sampling adequacy, the data sets were handled using the criteria suggested by Tabachnick and Fidell (2007) who stipulated .6 as the minimum value for a good factor analysis. The Bartlett’s test of Sphericity also needed to be significant (p <.05) for the factor analysis to be considered appropriate.
3.3.1.1 Constructs Created by Exploratory Factor Analysis
All questions with a Likert scale and their responses were subjected to individual Exploratory Factor Analyses (EFA). For example, all questions relating to the students’ attitudes towards whaling were contained within their own EFA. All questions that related to the objectives of the International Whaling Commission (IWC) were contained within their own EFA; likewise all questions relating to pro- whaling rhetoric were contained within their own EFA. Upon examination of the factor loadings73 produced by each EFA, questions that did not fall within finite groups were excluded from the overall construct creation74
A series of questions in the survey asked the participants to consider the importance of various goals during the formulation of whaling policy. These statements were
.
73
According to Tabachnick and Fidell (2007: 649), ‘the greater the loading, the more the variable is a pure measure of the factor’. Factor loadings in excess of .71 are considered excellent while loadings of .63 very good, .55 good, .45 fair, and .32 poor.
74
For example, upon examining the factor loadings produced by the questions in Table 3 (Chapter 4), all questions, excluding Question 11 (which produced a loading of -.828) produced factor loadings in the positive and therefore could be grouped together to form constructs.
Question 12 produced a low factor loading of .335, however including this question within the analysis when creating the approval of whaling construct was allowed because the Cronbach’s α value, needing to be >.7, was .88, indicating that the construct had a strong internal consistency (in accordance with DeVillis 2003).
also representative of the objectives of the IWC: ‘to provide for the proper conservation of whale stocks and thus make possible the orderly development of the whaling industry’. The questions that were factor analysed to determine the constructs based on the objectives of the IWC were survey questions 17i–vii (Appendix II).
Likewise, a number of questions were included in the survey that asked the participants to report their degree of acceptance of the rhetoric of pro-whaling advocates, in particular that produced by the Japanese Government. The associated questions that were factor-analysed were questions 6, 25i–vi and 25ix (Appendix II) and, based upon the groupings of their factor loadings, a construct or constructs were created.