Differential item functioning of a family affluence scale: Validation study on data from HBSC 2001/02

• Published in: Social Indicators Research
• Main Authors: Schnohr, C, Kreiner, S, Due, P, Currie, Candace Evelyn, Boyce, W, Diderichsen, F
• Format: Article in Journal/Newspaper
• Language: English
• Published: 2008
• Subjects: LOGLINEAR RASCH MODELS; SCHOOL-AGED CHILDREN; SOCIOECONOMIC-STATUS; CONSTRUCT-VALIDITY; ADOLESCENTS; COUNTRIES; MULTILEVEL; HEALTH; TESTS; DEPRIVATION; Canada; Currie; Greenland; Norway
• Online Access: https://research-portal.st-andrews.ac.uk/en/publications/43466e70-1497-4f7c-a389-cb442a35cfad; https://doi.org/10.1007/s11205-007-9221-4

Methodology for making cross-national comparisons is an area of increasing interest in social and public health related research. When studying socio-economic differences in health outcomes cross-nationally, there are several methodological issues of concern, especially when data is derived from self-reported questionnaires. Health Behaviour in School-aged Children (Currie et al. 1998) is a WHO cross-national study using school samples. HBSC provides comparable data, and thereby a unique opportunity to study associations between social indicators and health outcomes within an international perspective. In 2001/02 data was collected from a total of 162,323 children in 32 countries (Austria, Belgium, Canada, Croatia, Czech Republic, Denmark, Estonia, Finland, France, Germany, Greece, Greenland, Hungary, Ireland, Israel, Italy, Latvia, Lithuania, Macedonia, Malta, Netherlands, Norway, Poland, Portugal, Russia, Slovenia, Spain, Sweden, Switzerland, Ukraine, United Kingdom, USA.). Studies of social inequalities requires that a comparable measure of socio-economic position (SEP) is in use. HBSC has developed a proxy for social position measuring material wealth, the Family Affluence Scale (FAS). This paper studies FAS and whether it is comparable across population subgroups defined by country, age and gender. Initial analysis revealed that an item measuring perceived family wealth was a valid FAS item. Including this item in the FAS score will improve the reliability of FAS. Graphical log-linear Rasch models (GLLRM) showed that FAS contain differential item functioning (DIF) with respect to country, age, and gender as well as local dependency (LD) between items. During the analysis, test equating techniques where used to adjust for the test bias generated by DIF. We recommend that the equated scores are used whenever FAS is included as a variable. This study suggests that HBSC-FAS should contain five items (additional item: perceived family wealth) when analysing data from HBSC 2001/02, and furthermore that each ...