The data sgp package offers classes and functions to calculate student growth percentiles and percentile growth projections/trajectories using large scale longitudinal education assessment data. These analyses allow schools and districts to detect individual student achievement trends as well as understand how their students compare with those elsewhere.
An essential component of the SGP model is that academic peers are identified through state assessment scores over time. A student’s growth percentile can then be determined by comparing their current state assessment score against that of all those students from their grade and subject who previously took that same state assessment test; quantile regression provides this comparison tool; this considers other students who may have followed similar paths through assessment systems as the student themselves.
Fall 2024 will see teachers who qualify receiving three years’ of mSGP data. When assigning ratings to teachers, the NJDOE will use either the most recent year of data or median of previous two combined (whichever best serves educators).
The New Jersey Department of Education uses course roster submission and mSGP data to calculate teacher mSGP scores. Districts submit roster submissions during summer break after each school year containing lists of students assigned to qualifying teachers; NJDOE then links this mSGP data using score and subject code from each teacher’s data set, which results in its conversion into 1-4 scores that can then be tabulated alongside teacher practice scores and SGO.
The SGP package is built for use with R, an open source software environment available on Windows, OSX and Linux systems. Executing SGP analyses requires knowledge of R’s language as well as its associated tools and libraries; we advise those planning on performing such analyses to take time familiarizing themselves with R before diving in; most of the work in running SGP analyses lies in prepping data correctly for analysis; once this step has been accomplished executing analyses is typically straightforward and fast.