TÍTULO X DISPOSICIONES FINALES
CAPITULO ÚNICO
This section evaluates the analytical procedures adopted in terms of their adequacy to the attainment of the aim and objectives of this study. A quantitative analysis has been adopted to generate information and to uncover potential differences between the independent variables. Univariate, bivariate and multivariate analyses were adopted and are explained below.
4.11.1. Univariate and Bivariate Analysis
Univariate analysis (mean, median and frequency) was used to obtain descriptive information. Additionally bivariate analysis has been used, mainly to test the null hypothesis that there were no differences between the groups composing the independent variables. In this study, independent variables varied according to the objectives of the analysis, as explained in further detail in Chapter 4 (Methodology). When data was categorical, Chi-square for Independence was used. Nevertheless, when the minimum expected count requirement (e.g. expected frequency in each cell should be of 5 or more in a 2 by 2 table) was not sufficient the test results are not presented (Chapter 5, Section 5.4.2, Table 5.33). Although the respondents’ answers could have been grouped into fewer groups of answers to allow for the running the test, it could also contribute to potential loss of some of the specific information.
With ordinal data, the Mann-Whitney or Kruskal-Wallis tests were used. Given the characteristics of the data (lack of normality of distribution) and also the characteristics of non-parametric tests (have fewer assumptions) (Pallant 2007;
Field 2009), nonparametric tests were considered to be more appropriate in this research.
The criterion of 95% confidence, or a 0.05 probability was used in this research for rejecting the null hypotheses as has been accepted in research method literature as useful level for confidence (e.g. Pallant 2007; Field 2009). Thus, only when the level of confidence was 95% the hypothesis was not rejected (Field 2009). If a significance value of, for example, 0.01 was chosen, it would have reduced the probability of making a Type II error (when the null hypothesis is not rejected when, in fact, should have been). However, a lower significance value (0.05 as adopted in this research) provides a stronger indication for rejecting the null hypothesis and therefore reduced the probability of making a Type I error (null hypothesis is rejected when, in fact, it should have not have been rejected). In other words, by adopting a low significance value, the probability of making Type II
error was reduced. Subsequently, 0.05 was considered the appropriate significance level to adopt in this research.
However, and in the particular case of Kruskal-Wallis and when significant differences were found, Mann-Whitney U tests were performed to identify in which groups the differences were. In this case, and as indicated by Pallant (2010), a Bonferonni adjustment was applied. Thus, instead of considering p=0.05 as the significance level, a significance alpha of 0.017 (0.5/3), was used. The number 3 is the number of the Mann-Whitney U tests that would be done for the purposes of each analysis. In this study, in all the situations where Bonferonni adjustment was applied, 0.017 was always used as the criteria for determining if there were or not significant differences.
4.11.2. Multivariate analysis
Multivariate analysis, more specifically Hierarchical (agglomerative) Cluster Analysis was used in this study with regard to the personality variable. This technique allowed the identification and classification of three types of respondent (based on their personality traits). As a result of the analysis, it was possible to determine if the decision to cooperate in the future was related, or not, to the personality of the respondents.
Although the researcher had initially pondered doing a Principal Component Analysis to identify patterns of data and to reduce the number of dimensions without losing too much information, it was considered unnecessary given the deductive approach and that the content of the questions was informed by previous research identified in the literature review.
4.12 CONCLUSION
This chapter has presented the conceptual framework, the research aim, the objectives, the research questions and the underlying research hypothesis. It has also set out the research methodology and methods adopted in this study. Particular
attention was given to the research process, specifically questionnaire design, sampling procedures, and data analysis, namely the selection of the statistical procedures adopted. An evaluation of the methodology, methods and procedures considering its adequacy in the attainment of the aim and objectives was also provided.
This research has adopted a positivist stance, thus a quantitative analytical approach was adopted, with data being collected through a survey, based on an interview-based questionnaire filling. The research was conducted in the Douro Valley in the North of Portugal. In total, 200 (100 of tourism and 100 of wine respondents) questionnaires were obtained. Data was analysed with SPSS based on descriptive and inferential statistics. Different types of independent variables were taken into consideration, depending on the objectives of the analysis. This chapter has also provided an evaluation of methodology, methods, and analytical procedures that were adopted in this study in terms of their adequacy in the attainment of the aim and objectives of the study.
The next chapter presents behaviour-related findings of this study, focusing on the cooperation behaviour and experience of wine and tourism respondents in the past and the intention to whether cooperate, or not, in the future with other wine and tourism businesses in the future. Based on these findings, an emphasis will be given in the next chapter to cooperation (the characteristics and nature of cooperation) in the Douro Valley (in the context of wine and tourism industries).