Similarity structure analysis and structural equation modeling in studying latent structures: An application to the Attitudes Towards Portuguese Language Questionnaire

Abstract

Several international studies such as PISA and PILRS (Progress in International Reading Literacy Study), have stressed the importance of positive attitudes and behaviours as facilitators of individuals reading literacy during the school years and throughout their lives. Considering that there are not available instruments for assessing attitudes Towards Portuguese Language, it was proposed the development of the Attitudes towards Portuguese Language Questionnaire – ATPLQ (Questionário de Atitudes Face à Língua Portuguesa: QAFLP, Neto et al., 2011; Rebelo, 2012). The questionnaire has 22 Likert-type items, with four levels of response (Strongly Disagree, Disagree, Agree, Strongly Agree), spread, through exploratory factor analysis (EFA), over three attitudinal dimensions: Behavioural, Affective, and Motivational. In this study we aimed to analyse the ATPLQ’s latent structure with a pooled sample data of 1441 participants, applying similarity structure analysis (SSA) and confirmatory factor analysis of ordinal data (CFA). The SSA was carried out with Hudap in order to identify the structural properties of the questionnaire and to assess its adequacy in a Portuguese population. The CFA was carried out with LISREL in order to assure structural validity, i.e., accounting for factorial validity, but also for factors’ convergent and discriminant validity, and composite reliability. These psychometric features allowed the comparison of both the EFA derived model and the SSA derived model. We justify the selection of the SSA’s model, and we discuss the similarities between the results generated by SSA and LISREL procedures, highlighting their use in modeling constructs with ordinal indicators.