IPSIBAT   26217
INSTITUTO DE PSICOLOGIA BASICA, APLICADA Y TECNOLOGIA
Unidad Ejecutora - UE
congresos y reuniones científicas
Título:
Benefits of alignment for testing measurement invariance of cultural‐bound constructs: Cross‐cultural equivalence of the Attention‐Related Driving Errors Scale
Autor/es:
MONTES, S.; DA ROSA KRAUER, N; LEDESMA, R.D.; DONCEL P; BIANCHI A; PADILLA, JL; CASTRO, C.; OVIEDO‐TRESPALACIOS, O; BARRAGAN, D
Lugar:
Montreal
Reunión:
Conferencia; 11st Conference of The International Test Commission; 2018
Institución organizadora:
The International Test Commission
Resumen:
The Attention-Related Driving Errors Scale (ARDES) is intended to measure experiences distraction-related errors while driving. The Spanish source version was developed by Ledesma et al, (2010) in Argentina, then was adapted to Spain (Roca et al, 2013), UK (Peña-Suarez et al, 2016), US (Barragan et al, 2016) and currently adapted to Brazil and Australia. Ledesma et al. (2015) obtained validity evidence of three-factor structure for ARDES measures (Navigation, Manoeuvring and Control). Cross-cultural validity evidence of ARDES measures have not been obtained yet. Confirmatory Factor Analysis (CFA) approach to testing equivalence of ARDES measures can be too restrictive given the number of country groups, cultural and traffic rule differences, translation issues, etc. New strategies to test approximate equivalence like ?alignment? (Asparouhov & Muthén, 2014), can be more informative.ObjectivesTo test approximate measurement invariance and identify source of inequivalence for ARCES cross-cultural comparisons.Design/Methodology2040 participants responded to ARDES versions: Argentinian (603), Spanish (310), Brazilian (220), Australian (378), UK (298) and US (278) samples. The ARDES consists of 18 polytomous items. Alpha average is .84. ARDES total mean scores ranged from 1.46 (Argentina) to 1.73 (UK). A robust maximum likelihood fixed alignment is conducted to test approximate measurement invariance. Results10 intercepts and 17 item loadings out of 18 items are invariant across country groups. Eleven percent of the intercepts and 0.92% of the loadings are found non-invariant, averaging to 6% of non-invariance. ConclusionsAlignment results support approximate metric and scalar invariance across ARDES adaptations. We discuss on benefits of alignment approach to test equivalence when the number of groups is not large but there is cultural, linguistic and translation issues that can threat equivalence.