TITLE:
Psychometric Evidence of the 10-Item Connor-Davidson Resilience Scale (CD-RISC10, Greek Version) and the Predictive Power of Resilience on Well-Being and Distress
AUTHORS:
Theodoros Kyriazos, Anastasios Stalikas
KEYWORDS:
Resilience, CD-RISC10, Factor Analysis, Validation, Bifactor, CFA, EFA, SEM, Normative Data
JOURNAL NAME:
Open Journal of Social Sciences,
Vol.9 No.11,
November
30,
2021
ABSTRACT: The purpose of this study was
to evaluate the construct validity of CD-RISC10 in a sample of
1089 Greek adults of the general population. The CD-RISC10 factor structure was
evaluated first with EFA in a 20% subsample and
confirmed with CFA (CFA1) in a different 40% subsample. A cross-validation CFA followed (CFA2) in a third 40% subsample (i.e. of equal
power with CFA1). Model fit comparison using -2ΔLL difference test suggested a bidimensional structure but bifactor
ancillary measures indicated that multidimensionality was weak to exclude the
unidimensional structure. Full weak measurement invariance across gender for
this unidimensional model was successfully established in the entire sample.
Partial strong measurement invariance was established after freeing intercepts
of 2 items and partial strict after freeing the error variance of 1 item.
Internal consistency reliability (α) was equal to three different model-based
reliability calculations (CR) at adequate levels (.85), corroborating one
another, although CD-RISC10 was not tau-equivalent.
The average variance extracted was .37 to evaluate model-based
convergent validity. Convergent and discriminant validity were evaluated
further with correlation analysis with a resilience measure, life satisfaction,
affectivity, depression, anxiety, and stress with all associations to the
expected direction. The predictive validity of CD-RISC10 was evaluated with a
SEM model of resilience regressed on two higher-order latent factors of
subjective well-being (SWB) and psychological distress, yielding significant
strong positive and negative effects respectively. Male scored significantly
higher than females thus, normative data were calculated over the total sample
and also separately by gender.