Maddeye Tepki Verme Çabasının Süre İle Tanımlandığı Madde Tepki Kuramı Modelinin Parametre Kestirimlerine Etkisi

Abstract

This study examines the impact of item response time on ability and item parameters using an effort-moderated item response theory (IRT) model. PISA 2018 mathematics data for Türkiye were analyzed, categorizing student effort as solution behavior or rapid-guessing behavior based on NT10 and NT20 response time thresholds. Comparisons were made between standard IRT and effort-moderated IRT models. Findings revealed that effort-moderated IRT models, particularly the NT20 effort-moderated model with a higher proportion of low-effort responses, improved model-data fit and ensured item fit for items that did not align well with the standard IRT model. The NT20 effort-moderated model had the lowest -log-likelihood, AIC, and BIC values, demonstrating superior model-data fit. The largest discrepancies in item difficulty parameters between standard and effort-moderated IRT models were observed in easy item groups, while item discrimination parameters exhibited generally minor differences. Test information functions indicated similar measurement precision across all models for ability estimates. Although ability parameter estimates showed minimal differences between models, the low proportion of rapid-guessing behavior in the dataset may have influenced this result. Correlations between response time effort (RTE) and ability parameters were higher using the NT20 threshold, but the overall effect of test effort on ability estimation remained limited. This study highlights the importance of considering response time and effort to enhance the validity of measurement outcomes.