Canonical Correlation Statistical Analysis

53 For the third and fourth multiple regressions, certain cases were selected from the original sample for the analysis. Therefore, descriptive statistics for the variables are similar and variables meet assumptions of normality except for perceived value of the project which is similarly skewed -3 and kurtotic 13 Table 10. Table 10. Descriptive Statistics for Continuous Variables Used in Multiple Regressions for Individuals that Rated Ongoing Projects Variable Mean SD Skewness Kurtosis Perception of Fishing 0.53 0.30 -0.68 -0.68 Fatalistic Thinking 3.41 1.98 0.47 -0.99 Future Perception -1.39 1.24 -0.01 1.41 Investment - 9,000 baht 1.96 0.95 -0.77 -0.24 Investment - 110,000 baht 1.80 1.00 -0.41 -0.92 Tsunami damage to productivehousehold materials 2.61 1.96 0.14 -1.20 Participation in Current Projects 0.29 0.40 0.93 -0.77 Perceived Value of Current Projects 5.36 0.81 -1.41 2.25 Willingness to Partipate in Proposed Projects 0.75 0.33 -1.04 -0.11 Perceived Value of Proposed Projects 5.34 0.89 -3.01 13.83 In addition, because the sample is now smaller 226, interpretability is less reliable. However, the recommended sample size for an analysis with ten variables is 50 5 for each variable. Therefore, this sample size is also well above the lower limit suggested for this analysis 50 and, therefore, interpretability is still reasonable Harlow 2005. Table 10 shows descriptive statistics for the same variables used in these multiple regressions, including only cases where the individual rated the value of at least one ongoing project.

4.3.2. Canonical Correlation

54 This canonical correlation is used to examine the correlations between two sets of variables, post-tsunami perceptions attitudes toward the occupation of fishing, level of fatalistic thinking, perception of the future, spending choices if given 9,000 baht, spending choices if given 110,000 baht, damage to productive and household materials from the tsunami and perceptions of rehabilitation livelihood projects willingness to participate and perceived value of ongoing and proposed projects. The overall canonical correlation test was statistically significant, Wilks’ Λ = 0.897, F 12, 800 = 3.71, p 0.01, with a small effect size R 2 = 0.103. Of the two canonical correlations extracted from the data, the first function was statistically significant R C = 0.288, p 0.01, R 2 = 0.08 with a small multivariate effect size. This canonical function explains 81 of the shared variance in this set of variables. These results indicate the first pair of canonical variates were marginally related. There were four variables that exceeded the criteria for interpretability used in this analysis |.30| Harlow 2005. Attitudes toward the occupation of fishing 0.86, damage to productive and household materials from tsunami 0.61, and fatalistic thinking -0.35 were strongly associated with the first post-tsunami perceptions canonical variate 55 Table 11. 56 Table 11. Canonical Variate Loadings and Canonical Correlations for Post- Tsunami Perceptions and Perceptions of Proposed Livelihood Projects First canonical variate canonical loading 14 P OST - TSUNAMI PERCEPTIONS Perception of Fishing 0.86 Fatalistic Thinking -0.35 Future Perception -0.20 Investment - 9,000 baht 0.20 Investment - 110,000 baht 0.13 Tsunami damage to productivehousehold materials

0.61 P

ERCEPTIONS OF LIVELIHOOD PROJECTS Willingness to Participate in Proposed Projects 0.98 Perceived Value of Proposed Projects 0.10 C ANONICAL C ORRELATION 0.29 On the perceptions of proposed project canonical variate, willingness to participate shows a strong positive association 0.98. None of the redundancy coefficients were sufficiently high enough to be interpreted |.30| Harlow 2005. Positive attitudes about fishing, greater damage to productive and household materials and an individual that is less fatalistic is associated with greater willingness to participate in proposed projects.

4.3.3. Multiple Regressions to Follow-up on Canonical Correlations