Addi van Bergen

Chapter 3 52 Index selection Based on the results of the measurement properties analyses, we identified the best performing index. Generalisability of this index was subsequently examined by testing the items in the other datasets, where available. Statistical analysis Analyses were performed with SPSS 19.0. Group differences were tested with Pearson Chi Square test (categorical variables) or Anova F-test (continuous variables). Linear regression analyses were used to assess relationships between risk factors and social exclusion indices. RESULTS Characteristics of the study populations Table 2 shows the sociodemographic characteristics of respondents in the four cities. As can be seen in Table 2, risk groups for social exclusion such as persons of non- Western origin, lower educated persons and persons living in deprived neighbourhoods were well represented in all four samples. Significant differences were found between the samples with regard to sex, age, ethnical background, educational level and the proportion of individuals living in deprived neighbourhoods. The observed differences reflect demographic variation between the four cities and the degree of oversampling in difficult to reach groups. Table 2. Sociodemographic characteristics of the respondents per sample (unweighted). Total Amsterdam sample Rotterdam sample The Hague sample Utrecht sample p (N=19,658) (N=6,511) (N=5,127) (N=4,220) (N=3,800) Sex (male, %) 43.3 41.1 45.8 44.2 42.5 .000 * Age (mean, SD) 51.0 (19.1) 58.2 (20.0) 49.3 (17.6) 48.8 (17.6) 43.3 (16.9) .000 & Non-Western ethnic background (%) 20.4 19.1 23.4 24.8 13.7 .000 * Low educational level (%) # 16.1 19.8 15.8 14.8 11.5 .000 * Living in a deprived neighbourhood (%) 34.2 39.6 30.7 36.2 27.2 .000 * * The P values were obtained by using Pearson’s Chi Square analysis. & The P value was obtained by using One-way Anova F-test. # No education and primary school.