Parametrik ve Parametrik Olmayan Madde Tepki Modellerinin Kestirim Hatalarının Karşılaştırılması
Tarih
2020Yazar
Kıbrıslıoğlu Uysal, Nermin
Ambargo Süresi
Acik erisimÜst veri
Tüm öğe kaydını gösterÖzet
The purpose of this study is to compare parametric and nonparametric item response theory models under different ability distributions, difficulty distributions, item numbers, and sample sizes. MML, MAP, DC, EH, and NBE estimation procedures was used. Results were examined in 405 simulation conditions. Monte Carlo simulation method was used in data generation. Moreover, empirical data was also used to compare estimation methods. “irtoys” package in R was used in MLE and MAP estimations, and “mirt” package was used in DC and EH methods. For NBE estimations, an R code was developed by the researcher. Model comparisons was made based on MAE and RMSE values. Furthermore, ANOVA analysis was conducted to evaluate which manipulated factors are significantly related to outcomes. In the evaluation of real data sets, model fit statistics were examined. The results indicated that estimation methods, sample size, item number, distributional properties of latent trait, and some cross combination of them affected both item and person parameter estimates. On the other hand, the distributional properties of the difficulty parameter did not effect person parameter estimate while it affected item parameter estimates under latent trait nonnormality. When the latent trait was not normally distribured, DC provided the least estimation error. In general, NBE method produces the highest estimation error. The analysis of empirical data yielded similar results with the simulation study. DC and EH methods were the ones that provided minimum error. The distributional properties of ability estimated in empirical data differed explicitly between parametric and nonparametric estimation methods.
Bağlantı
http://hdl.handle.net/11655/22495Koleksiyonlar
Künye
Kıbrıslıoğlu Uysal, N. (2020). Parametrik ve parametrik olmayan madde tepki modellerinin kestirim hatalarının karşılaştırılması. (Doktora tezi). Hacettepe ÜniversitesiAşağıdaki lisans dosyası bu öğe ile ilişkilidir: