Effects of survey weights on the ordinal logistic regression modeling of perceived difficulty in reading comprehension from PISA 2018

• Main Authors: Gucyam, Ronald Gerbaud Evangelista, Razal, Flynt Bernardino, Tan, Mark Lenard Cunanan
• Format: text
• Language: English
• Published: Animo Repository 2022
• Subjects: Reading comprehension > Philippines; Regression analysis; Mathematics
• Online Access: https://animorepository.dlsu.edu.ph/etdb_math/6; https://animorepository.dlsu.edu.ph/cgi/viewcontent.cgi?article=1004&context=etdb_math
• Institution: De La Salle University

Recent literacy statistics reveal that Filipino students have poor reading comprehension. This study employed a Proportional Odds (PO) Ordinal Logistic Regression analysis in identifying significant factors that affect the perceived difficulty of 15-year-old Filipino students in reading comprehension using the Programme for International Student Assessment (PISA) 2018 data. However, PISA employed a complex sampling design, and it is an uncommon knowledge that ignoring it would lead to misleading inferences. Thus, the effects of survey weights in PO modeling were also investigated. Results showed that coefficients and standard errors were overestimated when weights were ignored, and the significant predictors differ between the unweighted and weighted model. In the weighted model, nonsignificant predictors were found to be the support shown by the teacher in test language lessons and teacher-directed instructions. Meanwhile, the attitude of the students towards school learning activities, their exposure to bullying, their sense of belongingness in school, the interest being displayed by the teacher, and the way of stimulating reading engagement by the teacher were found to be significant. Among these, their exposure to bullying (BEINGBULLIED), and their sense of belongingness in school (BELONG) were found to be the most significant contributors to students' perceived difficulty in reading comprehension of students.