Analysis Sequence

Description

Rationale

1

Data screening

To detect Outliers with Mahalanobis distance control.

2

Multivariate Normality Test with Multiple tests

To test the multivariate normality assumption with Mardia’s multivariate kurtosis and multivariate skewness test, Henze–Zirkler’s consistent test, Doornik-Hansen omnibus test, and Energy test.

3

Confirmatory Factor Analysis (CFA) of the study measures

To confirm structure in this sample.

4

Descriptive Statistics

To calculate means, medians, standard deviations, and reliability coefficients (see Kyriazos, 2017 ), i.e. alpha and greatest lower bound estimate (glb; Jackson & Agunwamba, 1977 ) for the latent variables of the measurement model. It holds glb ≥ α (Mair, 2018) .

5

Test the SEM measurement model fit and indicator reliability

To evaluate the fit of the measurement model since four out of five measures were partially used. Reliability of the observed variables was evaluated to reject the likelihood that the exogenous constructs are redundant or they have a multicollinearity problem.

6

Model-based Reliability and Validity Analysis of the latent variables in the measurement model

To evaluate the Composite Reliability (CR; Werts, Linn, & Joreskog, 1974 ; ω coefficient; McDonald, 1999 ) and Average Variance Extracted (AVE; Fornell & Larcker, 1981 ), Maximum Shared Variance (MSV) and Average Shared Variance (ASV), evidencing model-based convergent and discriminant validity.

7

Cross-validating model-based Discriminant Validity with additional methods

To cross-validate model-based discriminant validity with the Fornell & Larcker (1981) criterion and the HTMT Ratio of Correlation Method (Henseler, Ringle, & Sarstedt, 2015) .

8

Full measurement invariance of the SEM measurement model

To test if the measurement model has invariant factors, factor loadings, intercepts, and residuals across gender.

9

Test the full SEM model fit

To evaluate the structural model fit.

10

A priori & post hoc power analysis of the full SEM model with the RMSEA

To evaluate the sample required for 80% power to reject a wrong model. An alpha level of 0.05 was assumed with RMSEA misspecification 0.05 (MacCallum et al., 1996) .

11

Hypotheses testing (H1-H4)

Ten direct paths were specified.