Project Estimation of Students' Education Outcomes with Achievement Test Score "Plausible" Values and Missing Questionnaire Data Program John Martinson Honors College Research Programs Term Spring 2026 Status Accepted Research Area Education, Statistics Description Missing data values are a common problem in education research when child or adult participants decline to answer certain survey questionnaire or educational test items. When large-scale educational assessment datasets are prepared for distribution, missing values on the achievement test items are treated using multiple imputation, while missing values on the survey questionnaire variables in the same dataset are left missing. Existing methods for analyzing such education data include deleting observation rows with any missing value, which is likely to produce inaccurate results, or re-creating several complex imputation steps to “re-do” the data preparation, which is not user-friendly. In this research study, we propose a method to analyze large-scale international assessment data using maximum likelihood estimation that is broadly applicable and can be implemented in one step using popular statistical software packages (Mplus [Muthén & Muthén, 2025], R [R Core Team, 2026], or Stata [Statacorp, 2025]). We will evaluate the method by computer simulation and illustrate its use with an analysis of education data. We will provide software syntax examples to facilitate implementation of the missing data handling method by applied researchers. Supervisor Anne Traynor Mentor Type of work required The undergraduate researcher(s) will: (1) utilize reading, media, and discussion with experts to learn about statistical methods for handling missing data, (2) help to choose an existing large-scale education dataset for illustration of the proposed statistical method, (3) help to prepare the dataset for analysis (for example, by computing descriptive statistics, recoding variables, creating composite variables), (4) analyze the dataset, (5) contribute to interpreting the analysis results, and (6) meet bi-weekly with the principal investigator and/or an advanced doctoral student for planning.
Optional Activities: (7) Help with programming or running a computer simulation to test the method, and (8) Develop a poster for the Spring or Fall 2026 OUR Undergraduate Research Conference at Purdue. Websites Qualifications - All majors welcome. No prior research experience required! - Required: Interest in education research. Experience with programming in R or Stata statistical software. - Preferred: Statistics coursework. Credit Hours 0 Weekly Hours 0 (estimated)