Homepage for Yang Liu


Yang Liu, Ph.D. Assistant Professor
Department of Human Development and Quantitative Methodology 3942 Campus Drive University of Maryland College Park, MD 20742-1115 Office: 1230B Benjamin Building Tel: (301) 314-1126 Fax: (301) 314-9245 Email: yliu87@umd.edu Last updated: 07/04/2019
Bio
Yang Liu is currently an Assistant Professor in the Measurement, Statistics, and Evaluation program of the Department of Human Development and Quantitative Methodology at University of Maryland, College Park. His research focuses on the development of statistical methods for analyzing item response data, as well as the adaptation of measurement modeling to psychological, educational, and health-related research. [ Curriculum Vitae \| Preprints and presentations ]
Courses
▸	EDMS 623: Applied Measurement: Issues and Practices EDMS 623 is a graduate level introductory course to educational and psychological measurement. Classical test theory, generalizability theory, and item response theory are introduced, as well as the fundamental concepts of reliability and validity. [ Syllabus ]
▸	EDMS 787: Bayesian Inference and Analysis EDMS 787 is a graduate level course on Bayesian statistics. The course covers the fundamental rationale of Bayesian inference, Monte Carlo methods, and applications in educational measurement, statistics, and evaluation. [ Syllabus ]
▸	EDMS 646: General Linear Models I EDMS 646 is a graduate level course on General Linear models. The topics include multiple linear regression, ANOVA, ANCOVA, repeated-measures ANOVA, MANOVA, and mixed-effects models [ Syllabus ]
▸	PSY 10 (UC Merced): Analysis of Psychological Data PSY 10 is a lower-division undergraduate level course on non-calculus-based statistics. The topics include descriptive statistics, sampling distribution, confidence intervals, hypothesis testing, and simple linear regression. [ Syllabus ]
▸	PSY 212 (UC Merced): Item Response Theory PSY 212 is a graduate level course on IRT, developed in collaboration with my colleague Dr. Ji Seung Yang. The topics include the basics of IRT models, their estimation, model fit assessment, and applications. [ Syllabus ]

Research Projects
▸	Semiparametric Models for Item Responses and Response Times We developed semiparametric factor models for item responses and response times using the technique of cross-product regression splines. The proposed models are more flexible than existing parametric ones and can be applied to a broad range of assessment data. The work is in part supported by the NSF Grant SES-1826535. Publication: Liu, Magnus, & Thissen (2016) Collaborators: Drs. Brooke Magnus and David Thissen
▸	Fiducial/Objective Bayesian Inference It was found that fiducial/objective Bayesian inference could enjoy more favorable frequentist properties compared to gold-standard frequentist approaches such as maximum likelihood. We also studied the asymptotic expansion of fiducial/posterior quantiles and proposed general strategies to construct probability matching fiducial/posterior distributions, which leads to higher-order accurate confidence intervals. Publications: Liu, Hannig, & Pal Majumder (2019), Liu & Hannig (2016, 2017) Collaborators: Drs. Jan Hannig and Abhishek Pal Majumder
▸	Uncertainty Quantification in Item Response Theory (IRT) When the calibrated tests are reused in subsequent studies for the purpose of linking, scoring, or building explanatory models, it is often ignored that the estimated item parameters and factor scores are subject to various sources of error. It is important to investigate the fit of the original IRT model to the new data, modify the model if necessary, and adjust the statistical inference about item parameters and scores based on the updated model. Publication: Liu, Yang, & Maydeu-Olivares (2019), Liu & Yang (2018a, 2018b) Collaborators: Drs. Ji Seung Yang and Alberto Maydeu-Olivares
▸	Test Security and Cheating Detection Repeatedly using items in high-stake testing programs provides a chance for test takers to have knowledge of particular items in advance of test administrations. We proposed a predictive checking method to detect whether a person uses preknowledge on repeatedly used items, using information from secure items that have zero or very low exposure rates. Publication: Wang, Liu, & Hambleton (2017) Collaborators: Dr. Xi Wang
▸	Item Fit Diagnostics for IRT Models Goodness-of-fit at the item and item-pair level helps to determine how a poorly fitting IRT model can be improved. To this end, we proposed several piecewise model fit diagnostics with known asymptotic reference distributions, and illustrated their use with both simulated and real data sets. Publications: Liu and Thissen (2012, 2014), Liu and Maydeu-Olivares (2013, 2014), Maydeu-Olivares and Liu (2015) Collaborator: Drs. Alberto Maydeu-Olivares and David Thissen
▸	Patient Reported Outcome Measurement Information System (PROMIS) PROMIS, funded by the National Institutes of Health (NIH), is a system of reliable, valid, and flexible assessment tools that measure patient-reported health status. For more information, please visit http://www.nihpromis.org/.

Source Code
▸	R code that computes a second-order probability matching prior for the person parameter in unidimensional IRT models. For more details, please refer to Liu, Hannig, & Pal Majumder (2019)
▸	R code that implements restricted recalibration of IRT models. A brief tutorial and a simulated data example were also included in the archive. For more details, please refer to Liu, Yang, & Maydeu-Olivares (2019)
▸	Fortran code that implements fiducial estimation of logistic graded response models. The code was compiled with gfortran 4.8.4 (using the makefile provided in the archive); it also requires libgfortran and liblapack. An example data set (lsat7.dat) and the input file (lsat7.conf, in the format of the standard fortran 2003 namelist) were also included. For a detailed description of the sampling algorithm, please refer to Liu & Hannig (2017).
▸	R code that calculates bivariate diagnostics for unidimensional 2-parameter logistic (2PL) and graded response models using Mplus output. For more details, please refer to Liu & Maydeu-Olivares (2014).
▸	R code that calculates univariate and bivariate diagnostics for unidimensional 2PL and graded response models using the IRTPRO "-prm.txt" file. For more details, please refer to Maydeu-Olivares & Liu (2015).
Doctoral Students
▸	Chen Tian Chen joined EDMS as a doctoral student in 2018. Her research interest include computerized adaptive testing and multidimensional item response modeling.
▸	Shuangshuang "Reni" Xu Reni received a Master's degree from EDMS at 2015, and rejoined the program as a doctoral student in 2017. Her research interests include (but not limited to) item response and cognitive diagnostic models.
Links
UMD Department of Human Development and Quantitative Methodology UC-Merced Psychological Sciences UNC-CH Quantitative Psychology UNC-CH Department of Statistics and Operations Research
Psychometric Society National Council of Measurement in Education Society of Multivariate Experimental Psychology
UCLA Statistical Computing Resources Wolfram Alpha: Computational Engine Comprehensive R Archive Network Netlib Repository C++ Resourses Comprehensive TeX Archive Network