Exploring the Underlying Factor Structure of the Parent Reading Belief Inventory (PRBI): Some Caveats

Abstract

Research Findings: The present study explored the underlying factor structure proposed a priori by the developer of the Parent Reading Belief Inventory (PRBI: B. D. DeBaryshe, 1995) using a local independent sample. The PRBI was developed to assess maternal beliefs about reading aloud to children and was designed to measure attitudes, perceptions, and values about how children learn, the content of what they learn, as well as parental teaching efficacy. The PRBI is purported to have 7 underlying subscales and a total score. Analyses showed internal consistency estimates that were similar to those reported by the authors of the PRBI. Using confirmatory factor analysis, we found good fit for only 2 of the 7 dimensions (Reading Instruction and Resources) and poor fit for overall models for the entire PRBI scale when modeled using a 2nd-order factor, correlated factors, or a single general factor. Practice or Policy: Limitations and future research are discussed.

Notes

Note. Correlations for items in the same subscale are bolded.

Note. Environmental input has only two items, so a confirmatory factor analysis could not be tested for it. CFI = comparative fit index; SRMR = standardized root-mean-square residual.

^a The estimation of this model resulted in a negative estimated residual variance for Item 39, so the residual was fixed to zero.

¹We are not aware of this method having been described before in the literature. It is, however, roughly analogous to the approach often used in structural equation modeling of testing the measurement and structural models sequentially, allowing for sources of misfit to be more precisely located (Anderson & Gerbing, 1988). The primary difference between our approach and the Anderson and Gerbing (1988) approach is that our approach builds the model up by adding variables to the model, whereas the Anderson and Gerbing approach includes all variables from the beginning but builds the model up by adding parameter constraints.

Note. Items 39 and 41 were estimated to have negative residual variances in some submodels; all such models were reestimated with the items’ residual variances fixed to zero. CFI = comparative fit index; SRMR = standardized root-mean-square residual.

^a The submodel including Reading Instruction and Environmental Input was empirically underidentified and could not be estimated.

^b The submodel including Teaching Efficacy, Positive Affect, Reading Instruction, and Environmental Input was empirically underidentified and could not be estimated.