four groups representing the geographical area they inhabit: north-east; north-west; centre and south. Consumers living in northern Italy were expected to have a more sophisticated business culture than those living in the south. 4. 3. Analysis The statistical analyses were carried out using SPSS 18. 00 and AMOS, and are divided into two consecutive steps. First, using exploratory factor analysis of the data collected by the first and second part of the questionnaire, a scale of Responsible Attitude was constructed. We evaluated the scale for reliability and then conducted a confirmatory factor analysis to assess validity (Churchill, 1979, Mentzer and Flint, 1997). Subsequently, the various hypotheses were validated, analysing the significance of the mean in Responsible Attitude (an independent variable) among different demographical groups identified by the variables illustrated above (dependent variables). Responsible attitude scale development The goal of our first step in the analysis was to measure the level of Responsible Attitude claimed by each of the consumers surveyed. An exploratory factor analysis was applied using Keiser’s criterion and experimenting with different numbers of factors to find a satisfactory solution (Tabachnick and Fidell, 2008). The variables included in the factor analysis referred to the questions submitted to the 16
consumers through the above-mentioned telephone survey, and was developed on the basis of literature presented in the section above. Running the factor analysis, we first checked for multicollinarity among variables, looking at the determinant of the correlation matrix (0. 84) to ensure that it was greater than the generally accepted value of 0. 000001. We also checked for the Kaiser, Mayer and Olkin (0. 774) measure to evaluate the adequacy of the sampling and ensure we had a distinct and reliable factor. We conducted Bartlett’s test of sphericity to check the existence of relations between the variables we included in the analysis (p<0. 05). We verified whether the factor analysis would have been better with two or more factors, and the analysis of eigenvalue plot confirmed that the best option was to have only one component. This choice was also confirmed by the analysis of total variance that showed an eigenvalue of the second component that was less than one. Cronbach’s α was used as the measure of reliability of the responsible attitude scale, with a result of 0. 843. We conclude, therefore, that the Responsible Attitude scale shows adequate reliability. The assessment of validity of the scale was produced using a confirmatory factor model with maximum likelihood estimation. The final model displays acceptably fit of indices (χ 2 =123. 24 (15), p=. 002; GFI=. 91; CFI=. 98; RSMEA=0. 07), which meet the recommended levels for a model with a good fit (Hair et al. 2006). This indicates that the developed responsible attitude scale is valid.
You've reached the end of your free preview.
Want to read all 44 pages?