Chapter 4.2 - 4.
3: Validity, Reliability,
and Structural Model Testing
4.2 Validity Test Results
4.2.1 Convergent Validity Test Results
Figure 4.3 shows that each variable, namely Competence (X1), Independence (X2), Audit
Quality (Y), and Audit Fee (Z), has outer loading values greater than 0.70. This indicates that
the measurement for each construct is adequate. The following table presents the outer
loading results:
Table 4.10 Outer Loadings Result
Source: Processed SmartPLS 4 Output, 2025
4.2.2 Discriminant Validity Test Results
The discriminant validity test using the HTMT (Heterotrait-Monotrait Ratio) criterion is one
of the latest and most recommended methods for assessing discriminant validity in SEM-
PLS models. HTMT measures how well different constructs are truly distinct from each
other.
- HTMT value < 0.85 (or < 0.90 in some literature): Discriminant validity is met, indicating
empirical distinctiveness among constructs.
- HTMT value ≥ 0.85 (or ≥ 0.90): Discriminant validity is not met.
Table 4.11 Discriminant Validity Test Results
Source: SmartPLS 4 Output, 2024
Based on Table 4.11, all HTMT values between constructs are below 0.85. This implies that
each construct—Competence, Independence, Audit Quality, and Audit Fee—are empirically
distinct from one another, and the model is considered valid.
4.2.3 Reliability Test (Cronbach's Alpha)
The reliability test aims to measure the internal consistency of each variable. This is
assessed using Cronbach's Alpha and Composite Reliability. According to Ghozali (2016), a
variable is considered reliable if both Cronbach’s Alpha and Composite Reliability values
exceed 0.70.
Table 4.12 Reliability Test Results
Source: Processed SmartPLS 5 Output, 2025
4.3 Structural Model Test (Inner Model)
The structural model test aims to predict the causal relationships between latent variables
by evaluating the R-Square, Q-Square, and Goodness of Fit (GoF).
4.3.1 R-Square Test Result
Table 4.13 R-Square Test Result
Source: SmartPLS 4 Output, 2025
Table 4.13 shows that the R-Square value is 0.784. This indicates that the predictive power
of the structural model is strong, as the value exceeds 0.75. Thus, competence,
independence, and audit fee collectively explain 78.4% of the variance in audit quality.
4.3.2 Q-Square Test Result
Table 4.14 Q-Square Test Result
Source: Processed SmartPLS 5 Output, 2025
The Q-Square value is used to evaluate the predictive relevance of the model. A Q² value > 0
indicates good predictive capability. Based on the results, the constructs of Competence
(X1), Independence (X2), and Audit Fee (Z) all have Q² values > 0, which means they have
good relevance in predicting the dependent variable, Audit Quality.
4.3.3 Goodness of Fit (GoF) Test Result
The GoF test is used to assess the overall fit of the model in explaining empirical data. GoF
values range from 0 to 1, with 0.10 indicating small, 0.26 moderate, and 0.36 large fit
(Ghozali, 2015). The following AVE values were obtained:
- Competence: 0.612
- Independence: 0.642
- Audit Quality: 0.580
- Audit Fee: 0.622
- Moderation (Z*X1 and Z*X2): 1
R² = 0.784
AVE Mean = (0.612 + 0.642 + 0.580 + 0.622 + 1 + 1) / 6 = 0.7427
GoF = √(0.7427 × 0.784) = 0.6756
The GoF value of 0.6756 indicates that the model has a high overall fit and is considered
capable of explaining the empirical data effectively.
