Data SGP – Avoiding Errors in SGP Analysis

Data SGP is a powerful analysis tool for educators and district administrators that enables them to compare student growth percentiles (SGPs) across students, teachers, schools, districts and states. It serves as an objective way of measuring growth in student achievement while simultaneously helping educators identify individual student strengths and weaknesses as well as target instructional resources accordingly. Furthermore, district leaders can utilize Data sgp for student evaluation purposes while informing teacher evaluation systems.

SGP analyses transform raw student assessment scores into scaled scores that can then be compared with an average scaled score across students in the same grade and subject area, to identify whether or not an individual student has met or exceeded proficiency targets based on historical growth trajectories, while also projecting future achievement trajectories for all students, including projections of when each will attain proficiency along with projected growth rates in percentiles to reach that milestone.

These projections and trajectories are estimated using statistical methods employing latent achievement trait models and machine learning algorithms to minimize estimation error for these unobservable models. While accuracy of estimates is of paramount importance, various factors can impact results of SGP analyses and require district personnel to understand and take measures to minimize error.

Especially when analyzing longitudinal data, where teacher influence can have a great effect on student progress and therefore influence the conclusions drawn from an SGP analysis, it’s essential that users are familiar with all processes involved and any possible errors that may arise when running such analyses. The SGP package offers users several vignettes and documentations to aid their understanding of these analyses, along with any errors which might arise in running them.

For districts to avoid errors during SGP analyses, data preparation must be properly executed prior to beginning. Most errors related to data preparation can be attributed directly back to errors associated with analysis itself. The SGP package provides two exemplar data sets – sgpData and sgpData_LONG – designed to facilitate preparation of data for SGP analyses. These data sets include variables necessary for SGP analyses, such as unique student identifiers and grade/time assessments of each student. Furthermore, the sgpData_LONG set offers a lookup table called sgpData_INSTRUCTOR_NUMBER that allows districts to associate student test records with instructors via unique identifiers attached to instructors.

SGP was developed for use within the R statistical software environment and requires computers with R installed; available for Windows, OSX, and Linux with open source licenses available free-of-charge to educators worldwide. Before using R for SGP analyses it is strongly advised that educators take some time familiarizing themselves with its programming in order to maximize efficiency of analyses while decreasing chances of errors due to unfamiliarity.