• Keine Ergebnisse gefunden

Correspondence between Culture-Level and Individual-Level Clusters

3. Method

4.2 Cluster Analyses of Family Model Value Profiles

4.2.3 Correspondence between Culture-Level and Individual-Level Clusters

culture-level and individual-culture-level cluster solutions are equivalent, this should be reflected in the distribution of cluster memberships: adolescents from cultures belonging to the interdependent cluster in the culture-level cluster analysis should by majority belong to the interdependent cluster in the individual-level cluster analysis, and analogically for adolescent from cultures that belonged to the other two clusters in the culture-level analysis.

Furthermore, it was explored if there is a gender-specific distribution of cluster memberships.

A multinomial logistic regression analysis was employed using culture-level cluster membership (henceforth culture-level cluster), culture (nested in the former), and gender as predictors for the outcome variable individual-level cluster membership (henceforth individual-level cluster). First, a full factorial model was computed. The model test showed that this model was significant, and all main effects as well as the interaction effect of Culture x Gender was significant (see Table 22). For comparison purposes, a second model including only the main effects of culture-level cluster, culture, and gender was computed. This model showed a significant model fit, indicating a non-fitting model, but the AIC criterion was lower as compared to the saturated model indicating that the less complex main-effects-only-model should be preferred over the full-factorial model (see Table 22).

Table 22

Multinomial Logistic Regression for Individual-Level Cluster Membership

Model Full Factorial Main Effects

Effects Wald df p Wald df p

Culture-Level Cluster 586.80 4 <.001 712.72 4 <.001 Culture [C-L Cl.] 65.90 14 <.001 74.63 14 <.001

Gender 8.28 2 <.05 10.93 2 <.01

C-L Cl. x Gender 3.01 4 .56 Culture x Gender [C-L Cl.] 23.77 14 <.05

Criteria Overall Model Test (LR) Model Fit (LR) Model Comparison AIC R2 (U) Chi2 df p Chi2 df p Intercept only 1642.12

Full Factorial 251.51 .493 1466.60 38 <.001 saturated model Main Effects 245.36 .486 1436.76 20 <.001 29.84 18 <.05 Note. C-L Cl. = Culture-Level Cluster. The acronym in squared brackets signifies that the preceding effect is nested in the factor Culture-Level Cluster. AIC = Akaikes Information Criterion. R2 (U) = Nagelkerke’s Pseudo R2 (Reduction of Uncertainty). The bold printed model is selected as the best (and final) model according to the AIC. The factors culture-level cluster, culture, and gender were used for the definition of subpopulations.

For the main effect of culture-level cluster (in the main effects model), all single comparison across the levels of culture-level cluster on individual-level cluster were significant (see Table 23). Thus, all category contrasts of the predictor variable had a significant effect on all category contrasts of the dependent variable. In the following, the differences across culture-level clusters, culture, and gender with regard to individual-level cluster memberships are described (see Table 24 and Figures 12 and 13).

Table 23

Individual-Level Cluster Membership: Contrasts for Culture-Level Cluster Interdependence vs.

Emot. Interdependence Interdependence vs.

Independence Emot. Interdependence vs. Independence Culture-Level Cluster B SE Wald B SE Wald B SE Wald Interdependence vs.

Emot. Interdependence 1.94 .20 95.19*** 3.12 .39 62.57*** 1.18 .41 8.44**

Interdependence vs.

Independence 3.13 .38 68.87*** 5.44 .49 121.43*** 2.31 .42 30.23***

Emot. Interdependence

vs. Independence 1.19 .37 10.32** 2.32 .37 38.60*** 1.13 .24 22.80**

Note. Single contrast effects for the main effects of Culture-Level Cluster Membership on Individual-Level Cluster Membershipfrom the main effects model.

+ p < .10. * p < .05. ** p < .01. *** p < .001.

In cultures belonging to the cluster representing the family model of (total) interdependence in the culture-level analysis (Indonesia, India and South Africa) a large majority of adolescents also belonged to the individual-level cluster representing the family model of (total) interdependence (see Figure 12). Furthermore, in those cultures that belonged to the cluster representing the family model of emotional interdependence in the culture-level analysis (China, Turkey, and Israel) the majority of adolescents also belonged to the individual-level cluster representing the family model of emotional interdependence (see Figure 12). Finally, in those cultures belonging to the cluster representing the family model of independence in the culture-level analysis (France, Germany, Switzerland, Japan) the majority of adolescents also belonged to the individual-level cluster representing the family model of independence (see Figure 12). Thus, with regard to cluster-membership the results showed that the individual-level cluster analysis was in accordance with the results found in the culture-level cluster analysis. This was expected according to Hypothesis 7. The gender main effect revealed that more males than females belonged to the individual-level cluster representing the family model of (total) interdependence, and more females than males

belonged to the individual-level cluster representing the family model of emotional interdependence (see Figure 13).

Table 24

Individual-Level Cluster Membership: Culture, Culture-Level Cluster, and Gender Differences

Males Females All

Indep. Emot.

Interd. Interd. Indep. Emot.

Interd. Interd. Indep. Emot.

Interd. Interd.

Culture % % % % % % % % %

Indonesia 0.0 8.9 91.1 0.6 12.1 87.3 0.3 10.7 89.0 South Africa 0.8 18.0 81.1 3.6 17.9 78.5 2.5 18 79.5 India 4.1 20.3 75.7 2.6 19.1 78.3 3.3 19.7 77.0 Cluster:

Interdependence 1.7 15.8 82.5 2.3 16.4 81.2 2.1 16.1 81.8 China 6.2 51.9 41.9 7.3 65.5 27.1 6.9 59.8 33.3 Israel 26.1 44.9 29.0 16.8 46.2 37.0 20.2 45.7 34.0 Turkey 18.1 41.7 40.3 24.7 51.9 23.5 21.6 47.1 31.4 Cluster:

Emotional Interdependence

15.6 46.2 38.6 15.9 55.7 28.4 15.6 51.6 32.8 France 46.7 32.2 21.1 44.5 47.3 8.2 45.5 40.5 14.0 Germany 49.6 44.5 5.8 46.0 45.4 8.6 47.6 45.0 7.4 Switzerland 45.5 41.8 12.7 60.5 35.5 3.9 54.2 38.2 7.6 Japan 64.9 26.0 9.1 67.7 26.2 6.2 66.7 26.1 7.2 Cluster:

Independence 51.5 37.0 11.4 53.7 39.2 7.1 52.8 38.3 9.0 All 22.1 32.1 45.8 23.8 36.4 39.8 23.1 34.5 42.4 Note. Indep. = Independence Cluster. Emot. Interd. = Emotional Interdependence Cluster. Interd. =

Interdependence Cluster. Males n = 1106. Females n = 1460. All n = 2566.

Figure 12. Individual-level cluster membership across Cultures and Culture-Level Clusters.

Figure 13. Individual-level cluster membership: Gender differences.

0%

10%

20%

30%

40%

50%

60%

70%

80%

90%

100%

Independence Emot. Interdep.

Interdependence

0%

5%

10%

15%

20%

25%

30%

35%

40%

45%

50%

Independence Emotional Interd. Interdependence

Male Female

Multilevel Analysis to Estimate the Between-Culture Variance Explained by Culture-Level Cluster Membership. The results of the null model showed significant between-variance-components of the intercept of Category 1 (Independence vs. Interdependence) with u0j (1) = 6.099 (χ2 = 547.78, df = 9, p < .001) and of the intercept of Category 2 (Emotional Interdependence vs. Interdependence) with u0j (2) = 1.733 (χ2 = 523.96, df = 9, p < .001), indicating significant cross-cultural differences with respect to the outcome variable. In the model including the culture-level effects of the dummy variables representing culture-level cluster membership, all four culture-level effects were significant (see Table C-1 in Appendix C). In this model the between-variance-components were reduced to u0j (1) = 0.236 (χ2 = 40.42, df = 7, p < .001), and u0j (2) = 0.038 (χ2 = 20.14, df = 7, p < .01), but still significant.

Culture-level cluster membership accounted for 96.1 % of the between-culture variance for Category 1 [(6.099 - 0.236) / 6.099 = .961] and for 97.8 % of the between-culture variance for Category 2 [(1.733 - 0.038) / 1.733 = .978]. Therefore, about 97 % of the between-culture variance with respect to Individual-Level Cluster Membership was explained by culture-level cluster membership.