Dorien Bangma

46 | CHAPTER 3 fluency test (categories words starting with a ‘D’, ‘A’ and ‘T’; Schmand et al., 2008); 1 minute for each letter). Finally, mathematical reasoning and numeracy were assessed with the WAIS– IV Arithmetic (Wechsler, 2008, 2012). Data analyses Statistical Package for the Social Sciences 23 was used for data analyses. Effects of age. To determine the effects of age on the performance on the FDM tasks, thirteen hierarchical multiple regressions were performed on sample 1 using total scores of the FDM tasks as dependent variables. In the first step (method: enter) demographic variables that might be of influence on FDM were included as independent variables, i.e., gender, years of education, employment status and annual year income. Age in years was included as an independent variable in the second step (method: enter). Results of step two will be presented, unless otherwise stated. Normality was violated regarding the TDT total score; therefore, an arcsine transformation for percentage data was used for the TDT (i.e., 2 ∗ #$%&'()*+ ! /100 ; Cohen et al., 2003). When significant effects of age on FDM were found and a summary or total score was used as dependent variable (i.e., FCAI, FDMI, IBQ and IGT) the hierarchical multiple regression analysis was repeated including the different subscales or components as dependent variables. A p-value ≤ .05 was considered statistically significant. Internal and external validity. When significant effects of age on FDM were found, the internal and external validity of these results were examined. To determine the internal validity, bootstrap analyses were performed to control for overoptimism (Steyerberg et al., 2001, 2003). In the bootstrap resampling procedure, 1,000 random samples of the original sample size were drawn with replacement from sample 1. Bootstrap 95% Confidence Intervals (CI) were estimated for the regression coefficients (b-values) as a measure of accuracy of the original regression analyses. When the 95% CI of the bootstrap analyses did not include the value of zero, the internal validation was considered satisfactory. To examine the external validity of the significant results, the hierarchical multiple regression analyses, including bootstrap analyses for significant results, were repeated using sample 2. The b-values of the hierarchical multiple regression analyses and the bootstrap analyses in the second sample were compared to the 95% CI of the conducted bootstraps on sample 1 to evaluate the external validity of the found age effects (Bleeker et al., 2003; Steyerberg et al., 2003). When b-values of sample 2 were within the 95% CI of the bootstrap of sample 1, the external validity was considered to be sufficient. Mediating effect of cognition. To investigate a possible mediating effect of cognition on the relation between age and FDM, a four steps procedure by Baron and Kenny (1986) was followed. Step one is described above, i.e., the investigation of the relation between age and FDM. For step two, thirteen additional hierarchical multiple regression analyses (method: enter) were performed to determine the relation between cognition and participants’ performance in FDM tests. Dependent variables were the total scores of each FDM test. Independent variables were the participants’ performances on the different standard neuropsychological measures. When standard neuropsychological measures are found to significantly relate to aspects of FDM, additional hierarchical multiple regression analyses (step