An aMNLFA Approach to Measurement Invariance of the Strength and Difficulties Questionnaire

• In Event: Poster Session 04
  In Poster Session: PS 04 Topic - Developmental Psychopathology

Abstract

Empirical evidence has suggested that there are differences in rates of internalizing and externalizing problems in children across development, race, sex, and socio-economic status (e.g. Biel et al., 2015). For example, Black children and children from low income backgrounds are rated as expressing problem behaviors in the classroom more often than White and middle-income children (e.g. Bates & Glick, 2013; Brooks-Gunn & Duncan, 1997). However, many of these indicators rely on questionnaires leaving them open to potential issues of reporter bias. Testing and adjusting for bias in measurement is imperative to understanding the “true” substantive developmental progression of child adjustment across diverse populations. If items are endorsed discrepantly, based on rater or child characteristics (differential item functioning, DIF), this can lead to inference problems in the latent factor indicator. Adjustments are required to parse out this bias from substantive group differences. If indicators are biased, they can contribute to deleterious consequences that reinforce structural inequalities and discriminatory treatment for low-income and minority children (e.g. Achilles et al., 2007). The present study aims to address these issues by leveraging moderated nonlinear factor analysis (MNLFA; Gottfredson et al., 2019) to test measurement invariance associated with the widely used Strengths and Difficulties Questionnaire (SDQ; Goodman et al., 2000), as a function of teacher race, family income-to-needs ratio (INR), and child age, race, and sex.
Data comes from the Family Life Project (FLP; N=1,292)—a study of children living in low-income, rural contexts in Pennsylvania and North Carolina (Vernon-Feagans et al., 2013). Teachers reported on child internalizing and externalizing problems via the SDQ when children were in preschool, kindergarten, 1st, 2nd, 3rd, and 5th grade. Child sociodemographic information was reported by parents at the first visit and teacher race was self-reported at each time point (Table 1). MNLFA was used to test and adjust for measurement non-invariance as a function of person-level moderators to estimate person-specific factor scores that maintain a common scale across groups and developmental time. These scores simultaneously account or meaningful differences in the latent factor as well as differences due to differential item functioning (DIF) as a function of continuous and categorical covariates.
Preliminary findings suggest that there is minimal DIF in the SDQ (Figure 1). The internalizing subscale had two items (“nervous and clingy” and “many fears”) that demonstrated DIF as a function of age, suggesting that teachers were less likely to endorse this item for older children, holding the true underlying latent score constant. At the latent level, average differences between groups were found after adjusting for existing age DIF. Externalizing symptoms were higher for boys than girls and for Black children compared to White children. Children living in households with a lower income to needs ratio had higher internalizing and externalizing difficulties. Children’s likelihood of higher internalizing problems increased with age, and externalizing problems decreased with age. Testing for and adjusting for DIF via the MNLFA approach greatly improves our confidence in the validity of the SDQ to be used across demographic groups in rural American contexts.

Authors

• Alyssa R Palmer, University of Minnesota - Twin Cities
  Presenting Author

• Keira Bryn Leneman, University of Minnesota - Twin Cities
  Non-Presenting Author

• Meriah L. DeJoseph, University of Minnesota - Twin Cities
  Non-Presenting Author

• Robin Sifre, University of Minnesota
  Non-Presenting Author

• Clancy Blair, New York University
  Non-Presenting Author

• Daniel Berry, University of Minnesota - Twin Cities
  Non-Presenting Author