What People Want: Exercise and Personalized Intervention as Preferred Strategies to Improve Well-Being and Prevent Chronic Diseases

Figures

• 

  Figure 1. Role of a healthy lifestyle in determining well-being and prevention of chronic diseases.

• 

  Figure 2. Panels of bar plots of the subjects’ intentions toward lifestyle changes: (a) marginal distribution of the five intentions with absolute counts, percentages (computed out of N=2762 respondents), and their 95% confidence intervals (CIs). Significance level codes of the Z-test (with Bonferroni correction) in the pairwise comparisons: H0:pi≤pr against H1:pi>pr, with i≠r=1,…, 5: *** p < 0.001. In the pairwise comparisons: H0:pi≥pr against H1:pi<pr, with i≠r=1,…, 5: °°° p < 0.001. (b) Within-sex distribution of the five intentions, with conditional percentages computed within females and males, respectively. (c) Within-age-classes distribution of the five intentions, with conditional percentages computed within each age class. (d) Within-sex-by-age-classes distribution of the five intentions, with conditional percentages computed within sex for each age class. Significance level codes in panels (b–d) of the Z-test for the single “intention-by-sex” or “intention-by-age-class” association: ‡ p < 0.05, ‡‡ p < 0.01, and ‡‡‡ p < 0.001 (positive association); † p < 0.05, †† p < 0.01, and ††† p < 0.001 (negative association).

• 

  Figure 3. Classification tree of subjects’ intentions toward lifestyle changes. Each leaf node (i.e., terminal CT node) reports the following, in this order: the predicted class (in bold); the probability of each class conditioned on the CT rule (so that the sum of probabilities in each leaf node is 1; the probability in bold pertains to the predicted class and can be regarded as the precision of the prediction in that leaf node); the total count (n) and the percentage of subjects that are classified in that leaf node; and the class counts (with the count of the predicted class in bold). The validation measures (overall statistics and statistics by class) relative to the random classifier and the test sets are reported in the two tables inside the graph as means computed over, respectively, the 10,000 permutations of the randomly generated class labels and the 100 repetitions of the bagging procedure, along with 95% confidence intervals built with the percentile method. Significance level codes of the permutation test for comparing the CT built on the full set with the random classifier: ** p < 0.01 and *** p < 0.001 (Appendix A).

• 

  Figure 4. Classification tree of at least one target action against no action in which subjects intend to invest economically. The meaning of the quantities appearing in the graph is explained below in Figure 3. Significance level code of the permutation test for comparing the CT built on the full set with the random classifier: *** p < 0.001 (Appendix A).

• 

  Figure 5. Classification tree of the target action “investing in a medical specialist consultancy for a tailored lifestyle program” chosen by subjects among the 2464 indicating at least one action in which to invest economically. The meaning of the quantities appearing in the graph is explained below in Figure 3. Significance level code of the permutation test for comparing the CT built on the full set with the random classifier: *** p < 0.001 (Appendix A).

• 

  Figure 6. Classification tree of the target action “buying a gym/swimming pool membership” chosen by subjects among the 2464 indicating at least one action in which to invest economically. The meaning of the quantities appearing in the graph is explained below in Figure 3. Significance level codes of the permutation test for comparing the CT built on the full set with the random classifier: * p < 0.05 and *** p < 0.001 (Appendix A).