Addi van Bergen

Chapter 5 100 Gierveld [36], the SCP Social Exclusion Index [20] and Social Cohesion and Trust [22]. The internal consistency, internal structure, construct validity and generalisability were found satisfactory [24]. Migration background. In line with the Dutch standard definition, country of birth or, in case of second-generation immigrants, country of birth of the mother and/ or father, as registered in the municipal population registers, were used to define migration background. Quantitative data analysis Descriptive statistics Analyses were restricted to respondents of Surinamese, Moroccan and Turkish origin with native Dutch respondents as the reference group. In order to control for the stratified sampling design and selective non-response, we used SPSS Version 22 Complex Samples Likelihood tests for the descriptive analyses of the prevalence of SE. Sampling weights were calculated by Statistics Netherlands based on a linear model with 9 sociodemographic variables and their interaction terms [37]. The significance level α was set at 0.001 to reflect the large sample size. Structural validity To test whether the SEI-HS factor structure holds across the three migrant groups, we conducted confirmatory factor analyses in data subsets per migrant group using SPSS Amos 22.0. Five of the standard goodness-of-fit statistics given in Amos were used to assess model fit i.e. root mean square error of approximation (RMSEA), upper bound of 90% confidence interval (HI90), Tucker-Lewis index (TLI), comparative fit index (CFI) and Hoelter’s .05 Index [38]. The Chi square statistic was not considered given its sensitivity to large sample sizes. The model fit was considered good if RMSEA < 0.05, HI90) < 0.06, TLI ≥ 0.95, CFI > 0.90 and Hoelter’s .05 Index ≥ 200 [38]. These same criteria were used in the development of the SEI-HS [24]. Differential item functioning (DIF) DIF occurs when one group of individuals responds differently from another group on a given questionnaire item, even though both groups are equivalent on the underlying construct that is assessed, or in DIF terminology, if both groups show the same ability on the matching variable. In this study the categories ‘little or no’, ‘some’ and ‘moderate to strong’ of the relevant dimension scale were used as the ability levels. The cut-off points for these categories were based on the 85 th and 95 th percentile in the Dutch adult population of 19 years or older in 2012 [24]. For each immigrant group, three hierarchical models were calculated with SPSS ordinal logistic regression, with Y being the SEI-HS item tested, M the matching variable (i.e. the corresponding SE dimension) and G the grouping variable (i.e. Surinamese, Moroccan or Turkish versus Dutch):