Main analyses

4.2.1 Hypotheses one, two, and three

Following Thompson (1991), our strategy for interpreting the results of the canonical correlation analysis included two main steps. First, we evaluated the

significance and effect size of the full canonical model to determine whether or not an effect exists. Because we found a significant relationship in the first step, we

subsequently examined standardized weights and structure coefficients to understand each observed variable’s contribution to the relationship.

Step 1: Evaluation of full canonical model.

The bivariate correlations among the five observed variables are reported in Table 4-1. Correlations, standardized canonical coefficients are displayed in Table 4-2 and Figure 4-1. In the present analysis, the canonical model was significant, which is equivalent to the model including the two canonical correlations, χ ²(6) = 84.171, p

< .001, and with the first canonical correlation removed, χ ²(2) = 17.009, p < .001.

The first canonical correlation was significant, ρ1 = 0.514, a large effect size.

The second canonical correlation was also significant, ρ₂ = 0.273, a small effect size. In the first canonical variate pair, χ extracted 71.1% of variance of its own indicators

Table 4-1

Zero-order correlations between indicators in the canonical correlation model.

Variable 1 2 3 4 5

Note. * p < .05. MSCIT-BF = Multidimensional Scale of Chinese Individual Modernity-Brief Form; MSCIM-BF = Multidimensional Scale of Chinese Individual

Traditionality-Brief Form; ATSPPH-SF = Attitudes Towards Seeking Professional Psychological Help Scale-Shortened Form; ATIHS = Attitudes toward Indigenous Healing Scale.

Table 4-2

Correlations and standardized canonical coefficients for the first (χ) and second (η) canonical variates

First Canonical Variate (χ) Second Canonical Variate (η) Correlation Coefficient Correlation Coefficient Indicator set 1

Note. MSCIT-BF = Multidimensional Scale of Chinese Individual Modernity-Brief Form; MSCIM-BF = Multidimensional Scale of Chinese Individual Traditionality-Brief Form; ATSPPH-SF = Attitudes Towards Seeking Professional Psychological Help Scale-Shortened Form; ATIHS = Attitudes toward Indigenous Healing Scale.

(traditionality- modernity indicators), and explained 18.8% of variance of the opposite indicators (help-seeking attitudes indicators). η1 extracted 30.2% of variance of its own indicators (help-seeking attitudes indicators), and explained 8.0% of variance of the opposite indicators (traditionality-modernity indicators). In the second canonical variate pair, χ2 extracted 28.9% of variance of its own indicators (traditionality-modernity

Figure 4-1

Loadings and canonical correlations for the first and second canonical variate pairs.

Note. MSCIT-BF = Multidimensional Scale of Chinese Individual Modernity-Brief Form; MSCIM-BF = Multidimensional Scale of Chinese Individual Traditionality-Brief Form; ATSPPH-SF = Attitudes Towards Seeking Professional Psychological Help Scale-Shortened Form; ATIHS = Attitudes toward Indigenous Healing Scale.

indicators), and explained 2.2% of variance of the opposite indicators (help-seeking attitudes indicators). η₂ extracted 39.6% of variance of its own indicators (help-seeking attitudes indicators), and explained 3.0% of variance of the opposite indicators.

The significant correlations between the two canonical variates provide support for the first hypothesis. We were also interested in further exploring which specific help-seeking attitudes variables (e.g., ATSPPH-SF, ATIHS Value, and ATIHS Stigma) were related to traditionality and modernity. Our second hypothesis stated that

-.441

counseling attitudes, and our third hypothesis stated that modernity would be

significantly correlated with counseling attitudes, but not indigenous healing attitudes.

To test these hypotheses, we examined zero-order correlations between the independent and dependent variables, and evaluated which indicators contributed to the canonical relationships.

Step 2: Evaluation of indicators contributing to the relationships between canonical variates.

In analyzing which variables contributed to the canonical relationship, we used .30 as the cutoff point above which a specific observed variable would be

considered to be contributing to the canonical variate. Results from the first canonical variate pair indicate that lower MSCIT-BF (-.973) and higher MSCIM-BF (.690) are associated with lower ATIHS Value (-.418), lower ATIHS Stigma (-.608), and higher ATSPPH-SF (.603). The results of the second canonical variate pair indicate that higher MSCIM-BF (.724) is associated with higher ATIHS Value (.868), lower ATIHS Stigma (-.441), and higher ATSPPH-SF (.490).

Partial support was found for hypothesis two. As hypothesized, MSCIT-BF was significantly positively correlated with ATIHS Value (r = .26). In addition, in the first canonical variate, MSCIT-BF had a direct relationship with ATIHS Value scores.

However, contrary to expectations, MSCIT-BF was positively correlated with ATIHS Stigma (r = .28) and negatively correlated with ATSPPH-SF (r = -.27). Similarly, in the first canonical variate, MSCIT-BF was directly related to ATIHS Stigma, and inversely related to ATSPPH-SF.

Partial support was also found for hypothesis three. As hypothesized, MSCIM-BF was significantly positively correlated with ATSPPH-SF (r = .31) and not

significantly correlated with ATIHS Value (r = .02). However, contrary to expectations,

MSCIM-BF was significantly negatively correlated with ATIHS Stigma (r = -.30). In the first canonical variate, MSCIM-BF was directly related to ATSPPH-SF, and inversely related to ATIHS Value and ATIHS Stigma. In the second canonical variate, MSCIM-BF was directly related to ATSPPH-SF and ATIHS Value, and inversely related to ATIHS Stigma. Thus, in both zero-order and canonical correlation analyses, MSCIM-BF was directly related to ATSPPH-SF, and inversely related to ATIHS Stigma. MSCIM-BF had no clear relationship with ATIHS Value, due to mixed relationships in the canonical correlation and an insignificant zero-order correlation.

4.2.2 Hypothesis four

We tested the multiple mediation model using lavaan. The multiple mediation model is depicted in Figure 4-2. Our analysis included three steps. First, we tested the significance of the overall mediating effect by adding up all 36 possible mediating effects from two IVs to three DVs through the six proposed mediating variables. lavaan was unable to calculate values for paths kdd and ldd, possibly due to the very weak correlation between DDI and ATIHS Stigma (r = -.02). Thus, when calculating mediation effects involving kdd or ldd, we simply removed those paths from the calculations. The global mediation effect was 0.148 with a standard error of 0.048 (Z = 3.077, p = 0.002), indicating a significant mediation effect for the overall model (Table 4-3).

Table 4-3

Test of the overall mediating effect from two IVs to three DVs.

Label Est SE Z p

1 Overall effect

0.148 0.048 3.077 0.002**

Note. ** p < .01.

Figure 4-2

Indicators and path names in the multiple mediation model.

Note 1. Each of the lines above corresponds to a path between two variables. Mediation effects are represented by a pair of paths connecting at a mediating variable. For

example, the mediating effect between MSCIT-BF and ATSPPH-SF through SNHS is represented by the paths labeled "a" and "m," is labeled as "am," and is calculated as a*m.

Note 2. MSCIT-BF = Multidimensional Scale of Chinese Individual Modernity-Brief Form; MSCIM-BF = Multidimensional Scale of Chinese Individual Traditionality-Brief Form; SSRPH = Social Stigma for Receiving Psychological Help Scale; SSOSH = Self-stigma of Seeking Help Scale;DDI = Distress Disclosure Index; SNHS = Subjective norm for help-seeking; ATSPPH-SF = Attitudes Towards Seeking Professional Psychological Help Scale-Shortened Form; ATIHS = Attitudes toward Indigenous

MSCIM-BF

We then tested the significance of the meditating effect of the set of six

mediating variables on each of the six relationship pairs between the two IVs and three DVs (Table 4-4). First, we tested the significance of the overall mediation effect of the six mediating variables on the relationship between MSCIT-BF and ATSPPH-SF. The estimated mediation effect was -0.015 with a standard error of 0.019 (Z = -0.827, p

= .408), indicating a non-significant effect. Second, we tested the significance of the overall mediation effect of the six mediating variables on the relationship between MSCIT-BF and ATIHS Value. The estimated mediation effect was 0.028 with a

standard error of 0.017 (Z = 1.624, p = 0.104), indicating a non-significant effect. Third, we tested the significance of the overall mediation effect of the six mediating variables on the relationship between MSCIT-BF and ATIHS Stigma. The estimated mediation effect was 0.023 with a standard error of 0.007 (Z = 3.186, p = 0.001), indicating a significant effect. Fourth, we tested the significance of the overall mediation effect of the six mediating variables on the relationship between MSCIM-BF and ATSPPH-SF The estimated mediation effect was 0.085 with a standard error of 0.024 (Z = 3.565, p

< .001), indicating a significant effect. Fifth, we tested the significance of the overall mediating effect of the six mediation variables on the relationship between MSCIM-BF and ATIHS Value. The estimated mediating effect was 0.043 with a standard error of 0.023 (Z = 1.905, p = 0.057), indicating a non-significant effect. Sixth, we tested the significance of the overall mediation effect of the six mediating variables on the relationship between MSCIM-BF and ATIHS Stigma. The estimated mediating effect was -0.016 with a standard error of 0.010 (Z = -1.693, p = 0.090), indicating a non-significant effect.

Finally, because we found two sets of significant mediating effects (between MSCIM-BF and ATSPPH-SF and between MSCIT-BF and ATIHS Stigma), we

Table 4-4

Tests of mediation effects from each IV to each DV.

Label Est SE Z p

Tests of individual mediation effects.

Label Est SE Z p

MSCIT-BF to ATIHS Stigma (ayczeaagbbicc)

1 ay 0.000 0.001 0.411 0.622

2 cz 0.011 0.005 2.343 0.019*

3 eaa -0.001 0.002 -0.435 0.663

4 gbb 0.002 0.002 0.932 0.352

5 icc 0.010 0.005 2.148 0.032*

MSCIM-BF to ATSPPH-SF (bmdnfohpjqlr)

1 bm 0.029 0.016 1.848 0.065

investigated the specific mediating effects of the individual paths in this set (Table 4-5).

Two mediation effects in the relationship between MSCIT-BF and ATIHS Stigma were significant. For SSRPH as a mediator, the estimated mediation effect was 0.011 with a standard error of 0.005 (Z = 2.343, p = 0.019), indicating a significant effect. For

SSOSH as a mediator, the estimated mediation effect was 0.010 with a standard error of 0.005 (Z = 2.148, p = 0.032), also indicating a significant effect.

Two significant mediation effects were also found for the relationship between MSCIM-BF and ATSPPH-SF. For Anticipated utility as a mediator, the estimated

mediation effect was 0.023 with a standard error of 0.010 (Z = 2.159, p = 0.031), indicating a significant effect. For SSOSH as a mediator, the estimated mediation effect was 0.021, with a standard error of 0.010 (Z = 2.044, p = 0.041), also indicating a significant effect.