The significance of the indirect effects and the contrasts were determined via bootstrapping, a computationally extensive procedure (bootstrap samples: 5000) generating an empirical based estimation of the sampling distribution, which in turn is used for calculating bias corrected confidence intervals (CI)15. If a CI is completely below or above zero, the respective effect is significant15. Based on theory2, we controlled for demographics (age), functional health (physical quality of life) and illness type (cancer site; included in all analyses among the total sample). Additionally, we controlled for the visibility of the disease21. Gender was highly confounded with cancer type (prostate: all male; breast: all female except 3 patients) and thus not included. Robustness of effects was verified by replicating the findings across samples and reporting 95 % confidence intervals33. Multicollinearity was checked for each model (variance inflation factors ≤ 2.2). Multivariate outliers, detected via Mahalanobis distance exceeding the critical value of the chi-square-distribution with p < .001, were rare (≤ 1.4 %) and kept in the samples. Alpha was two-sided and set at .05. Listwise deletion was applied for all analyses, which were performed by SPSS 24 and the macro PROCESS32.
The significance of the indirect effects and the contrasts were determined via bootstrapping, a <br>computationally extensive procedure (bootstrap samples: 5000) generating an empirical based estimation of the sampling distribution, which in turn is used for calculating bias corrected confidence intervals (CI)15. If a CI is completely below or above zero, the respective effect is significant15. Based on theory2, we controlled for demographics (age), functional health (physical quality of life) and illness type (cancer site; included in all analyses among the total sample). Additionally, we controlled for the visibility of the disease21. Gender was highly confounded with cancer type (prostate: all male; breast: all female except 3 patients) and thus not included. Robustness of effects was verified by replicating the findings across samples and <br>reporting 95 % confidence intervals33. Multicollinearity was checked for each model (variance inflation factors ≤ 2.2). Multivariate outliers, detected via Mahalanobis distance <br>exceeding the critical value of the chi-square-distribution with p < .001, were rare (≤ 1.4 %) <br>and kept in the samples. Alpha was two-sided and set at .05. Listwise deletion was applied for all analyses, which were performed by SPSS 24 and the macro PROCESS32.
正在翻译中..