Unique Contributions or Measurement Error Applying a Bifactor
Unique Contributions or Measurement Error? Applying a Bi-factor Structural Equation Model to Investigate the Roles of Morphological Awareness and Vocabulary Knowledge in Reading Comprehension Michael J. Kieffer Yaacov Petscher New York University Florida Center for Reading Research
What I’m Not Talking About Kieffer, M. J. & Box, C. D. (2013). Derivational morphological awareness, academic vocabulary, and reading comprehension in Spanish-speaking language minority learners and their classmates. Learning and Individual Differences, 24, 168 -175.
What I’m Not Talking About Kieffer & Box (2013): • Inspired by Nagy, Berninger, & Abbott (2006) • Derivational MA made a direct unique contributed to reading comprehension, controlling for word reading fluency and academic vocabulary. • Derivation MA made indirect contributions to reading comprehension via both word reading fluency and academic vocabulary. • Predictive relations were largely similar for native English speakers and Spanish-speaking language minority learners
Morphological Awareness (MA) • Students’ metalinguistic understanding of how complex words are formed from smaller units of meaning • Starts as a oral language skill, but is developed through interaction with oral and written language • Develops throughout the grades, with derivational MA becoming particular important in upper elementary & middle grades (e. g. , Carlisle, 1995; for a review, see Kuo & Anderson, 2003)
What we already know Or at least think we know…
Morphological Awareness (MA) predicts Reading Comprehension (RC) MA RC • For a while, we have known that MA is correlated with reading comprehension (e. g. , Carlisle, 2000; Freyd & Baron, 1982; Tyler & Nagy, 1990)
MA predicts RC, above & beyond Vocabulary (V) MA RC V • Unique contributions of MA to RC, controlling for vocabulary (e. g. , Carlisle, 2000; Kieffer, Biancarosa, & Mancilla-Martinez, in press; Kieffer & Lesaux, 2008, 2012; Kieffer & Box, 2013; Nagy, Berninger, & Abbott, 2006)
MA V But wait… Are we actually measuring MA and vocabulary as different constructs? • Conceptually, MA and vocabulary knowledge both involve meaning units (e. g. , Kuo & Anderson, 2006). • Operationally, measuring MA requires meaningful manipulation of meaning units (e. g. , Carlisle, 2012)
MA V But wait… Are we actually measuring MA and vocabulary as different constructs? • Empirically, MA correlates moderately to strongly with vocabulary (Deacon, Wade-Woolley, & Kirby, 2007; Deacon, 2011; M J Kieffer & Lesaux, 2008; Mahony, Singson, & Mann, 2000; Pasquarella, Chen, Lam, Luo, & Ramirez, 2012; Ramirez, Chen, Geva, & Kiefer, 2010; Singson, Mahoney, & Mann, 2000; Wang, Ko, & Choi, 2009; Wang, Yang, & Cheng, 2009 – Some observed correlations above. 60 (Carlisle, 2000; Ku & Anderson, 2003; Wang, Cheng, & Chen, 2006 )
MA V But wait… Are we actually measuring MA and vocabulary as different constructs? • Observed correlations between MA and vocabulary are attenuated by measurement error • Reliability of researcher-created MA measures has been moderate – In the. 70 -. 80 range & occasionally lower • So, “unique” contributions of MA beyond V could be an artifact of measurement error
MA/V Reason to worry… • Using Confirmatory Factor Analysis (CFA), Muse (2005) found that MA could not be distinguished from vocabulary in fourth grade (See also Wagner, Muse, & Tannenbaum, 2007). • Spencer (2012) replicated this finding with eighth graders.
MA V On the other hand… • Using CFA, Kieffer & Lesaux (2012) found that MA was measurably separable from two other dimensions of vocabulary in Grade 6 – though they are strongly related • Neugebauer, Kieffer, & Howard (under review) replicated this finding for Spanish- speaking language minority learners in Grades 6 -8
Research Question 1 To what extent do morphological awareness and vocabulary knowledge constitute measurably separable dimensions of lexical knowledge in sixth grade?
Sample • 148 sixth graders in 2 suburban schools in Arizona Asian/Pacif ic Islander 2% Multiethni White c 20% 3% African. American 11% Latino 64%
Sample • Schools reported 81% and 65% of students receiving free or reduced lunch • 9% designated as English language learners
Measures • Derivational Morphological Awareness – Nonword suffix choice task (e. g. , Nagy et al. , 2006) – The man is a great ____. A) tranter B) tranting C) trantious D) trantiful – 18 items; Cronbach’s Alpha =. 78 • Vocabulary – Multiple-choice synonym task based on Lesaux & Kieffer (2010) – Words drawn from the academic word list (Coxhead, 2000) – 18 Items; Cronbach’s Alpha =. 74
Measures • Reading Comprehension – Gates-Mac. Ginitie Reading Test, 4 th Ed. (Mac. Ginitie, Maria, & Dreyer, 2000) • Control: Word Reading Fluency – Test of Silent Word Reading (Mather, Hammill, Allen, & Roberts, 2004 dim|how|fig|blue
Research Question 1: Data Analyses • Using item-level data for MA & Vocabulary • To investigate dimensionality: – Parametric exploratory factor analysis – Nonparametric exploratory factor analysis – Parametric CFA – Nonparametric CFA
Modeling Dimensionality of Lexical Knowledge: Unidimensional • Fit poorly • Rejected across parametric & nonparametric EFA & CFA models MA 1 MA 2 MA 3 … MA 18 V 1 V 2 V 3 … V 18 Lexical Knowledge
Modeling Dimensionality of Lexical Knowledge: Two Dimensional MA 1 MA MA 2 MA 3 … MA 18 . 78 V 1 Vocab V 2 V 3 … V 18 • Fit better • Latent factors were strongly related
Modeling Dimensionality of Lexical Knowledge: Bi-factor Model MA 1 MAspecific • Fit the best MA 2 MA 3 … MA 18 Lexical Knowledge V 1 Vocabspecific V 2 V 3 … V 18 CFI =. 98; TLI =. 98; RMSEA =. 015 >1 D: Δχ² = 66. 71, Δdf = 34, p <. 001 >2 D: Δχ² = 48. 94, Δdf = 33, p <. 05
Findings: Dimensionality • Morphological Awareness and vocabulary are measurably separable constructs – At least with these measures and in this population • A bi-factor model that accounts for both the overlapping construct of lexical knowledge and the uniqueness of vocabulary and morphological awareness fits best
Research Question 2 To what extent does morphological awarenessspecific variance uniquely predict reading comprehension, beyond vocabulary-specific variance and the common variance shared by morphological awareness and vocabulary knowledge in sixth grade?
Research Question 2: Data Analyses • Structural Equation Modeling using a bi-factor model to predict reading comprehension performance
Predicting Reading Comprehension MAspecific Lexical Knowledge. 10. 21 Vocabspecific Word Reading Fluency . 67*** Reading Comp
Findings: Predicting Reading Comprehension • Individually, each of MA-specific variance, vocabulary-specific variance and lexical knowledge strongly predicted reading comprehension. • Together, only lexical knowledge had a unique significant association with reading comprehension.
Discussion • Good news: Results support the common assumption that our measures are capturing different constructs. – But we need to keep collecting validity data anyway. • Bad news: What’s unique about MA did not uniquely predict reading comprehension beyond what it shares with vocabulary. – Maybe the unique contribution of MA is less robust than we think.
Limitations & Future Research • We accounted for item-level measurement error, but not task-level measurement error. • Statistical power was limited to detect small effects. • Small number of ELLs prevented analysis of measurement invariance.
Thank You For more information, email: michael. kieffer@nyu. edu
- Slides: 29