Using new data from large-scale assessments such as NAEP and TIMSS is difficult because simply merging the data and using the well-known plausible value and survey methodology will yield biased results. To help researchers with this problem, the EdSurvey product development team recently published a new version of Dire, an R package intended to complement its primary statistical software application. Dire, which is named for its direct estimation functionality, allows researchers to conduct linear regressions with a latent outcome variable. More specifically, the package implements a survey-weighted marginal maximum estimation, a type of regression where the outcome is a latent trait such as student ability. Instead of using an estimate, the likelihood function marginalizes the latent trait. The package also includes a variety of variance estimation strategies. Read more about weighted Marginal Maximum Likelihood (MML) estimation for student test data in the Dire package here.
For researchers who are unfamiliar with EdSurvey, Dire is part of a suite of R packages developed to enhance education researchers’ ability to conduct sophisticated analyses of NCES data from assessments such as NAEP. Please stay tuned for updates on the team’s upcoming trainings to see how you can use Dire in your work. If you are eager to get started using the EdSurvey package, access current trainings on the NCES Distance Learning Dataset Training website.