N.

Description

Rationale

1

Data screening

To detect outliers with Mahalanobis distance critical value.

2

Univariate normality evaluation with multiple tests

To test for skewness, kurtosis and the univariate normality assumption with Kolmogorov-Smirnov (Lilliefors), Shapiro-Wilk, Shapiro-Francia, and Anderson-Darling tests.

3

Multivariate normality evaluation with multiple tests

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

4

Sample-splitting (20%, 40%, 40%)

To carry out EFA (20%), an initial CFA1 (40%) and cross-validating CFA2 (40%) in three different subsamples, the sample was randomly divided into three parts (20%, 40%, 40%). The two CFA subsamples (40%) were of equal power (3-Faced Construct Validation Method, Kyriazos, 2018a ).

5

Exploratory Factor Analysis (EFA)

To establish a structure for CD-RISC10. The number of factors to retain was examined with Parallel Analysis (Horn, 1965) , Very Simple Structure (Revelle & Rocklin, 1979) , Minimum Average Partial Correlations (Velicer, 1976) and Bayesian information criterion (BIC).

6

Confirm the EFA results with a Confirmatory Factor Analysis (CFA 1)

The EFA structure of CD-RISC10 was confirmed with an initial CFA, to test alternative models including a bifactor model.

7

Tests of fit difference

To compare the model fit of all the alternative CFA1 models with the likelihood ratio test (−2ΔLL MLR rescaled version; Satorra & Bentler, 2010 ).

8

Evaluating the Bifactor model

To evaluate the bifactor model, using bifactor ancillary model fit measures (Reise, Bonifay, & Haviland, 2013) .

9

Evaluating the influence of outliers on CFA1

To test if outliers influenced CFA1 model fit with 2 trial CFAs (in a subsample with vs without multivariate outliers; Tabachnick & Fidell, 2013 ).

10

Cross-validating the CFA1 optimal model with CFA2

To cross-validate the optimal CFA1 in a different subsample of equal power (CFA2).

11

A priori & post hoc power analysis of the CFA2 model

To evaluate the sample required for achieving a power of 80% to reject a wrong model. An alpha level of .05 was assumed with an RMSEA misspecification of .05 (MacCallum, Browne, & Sugawara, 1996) .

12

Measurement invariance across gender to the strict level of the CFA2 model

To test if the cross-validated CFA2 model was invariant factors, factor loadings, intercepts, and error variances across gender.

13

Internal Consistency Reliability, Model-Based Reliability and Model-based Convergent Validity after testing tau-equivalency of the optimal model

To evaluate Cronbach’s alpha [95% CI], and the greatest lower bound estimate (glb; Jackson & Agunwamba, 1977 ). To evaluate Composite Reliability (CR; Werts, Linn, & Karl, 1974 ) with the standardized loadings using 3 calculations (Bollen, 1980; Bentler, 1972; McDonald, 1999) and Average Variance Extracted (AVE; Fornell & Larcker, 1981 ), evidencing model-based reliability (see Mair, 2018 ) and convergent validity respectively (see Hoque et al., 2017 ).