Since Masters proposed a Rasch family model (i.e., partial credit model, PCM) in 1982, so far, PCM had become the basic model for dealing with polytomous survey data (e.g., the Likert-type rating scale) and performance assessment data (e.g., the partial-credit scoring) in social and behavioral sciences. Thereafter, Yu (1991) in his dissertation proposed a two-parameter partial credit model (TPPCM), which expanded the one-parameter Rasch-type PCM to add the step discrimination parameter ( ) for each step j within the item i to become the TPPCM. In the meantime, Muraki (1992) proposed the generalized partial credits model (GPCM), which allowed each item shared a common discrimination parameters ( ) in PCM.
Due to Bayesian computing is a flexible method for data analysis and free software like R-language is becoming more and more popular in statistical modeling. Thus, in this current study, we try to explore and rebuild the GPCM and TPPCM models through the Bayesian framework with the probabilistic programming languages (PPL) (e.g., JAGS program) (Plummer, 2017) and use it on the R interface rjags package (Plummer, 2022).
Eight items that measured social competence are selected from KIT dataset (KIT, National Longitudinal Study of Child Development and Care) in Taiwan. 570 subjects are randomly sampled as the sample real data and used to implement the GPCM and TPPCM for checking and comparing their convergence, estimation efficiency, and inferences in this current study. Final results are proposed. The implications for future meaning and usage in social and behavioral sciences are also suggested.

Authors:
Jie-Wen Tsai, National Chengchi University, Taiwan
Min-Ning Yu, National Chengchi University, Taiwan

About the Presenter(s)
Mr Jiewen Tsai is a University Doctoral Student at National Chengchi University in Taiwan

See this presentation on the full schedule – Friday Schedule

Bookmark