The data sgp package provides classes, functions and exemplar data for performing student growth percentiles (SGP) calculations. SGP calculations utilize large scale, longitudinal education assessment data to create conditional density models that represent a student’s achievement history. These models are used to generate a set of percentile growth projections/trajectories that show how much a student needs to grow to reach future achievement targets.
SGPs compare a student’s growth trajectories to those of their academic peers statewide. This comparison allows teachers and districts to see if their students grew more or less than other students and identify areas for improvement.
In addition to providing the ability to compare growth trajectories to those of students in their classrooms, the state’s SGP score also gives educators an opportunity to identify and target groups of students who need the most help in order to improve their scores. To do this, schools and districts use SGPs in the context of a data driven instructional framework called the Performance Framework.
A guiding principle of the performance framework is that students must learn at different rates and in a variety of ways to make progress toward mastery. This is why the performance framework requires a balance of multiple measures of student learning, including both the traditional high stakes tests and non-test indicators of student learning such as grades and school-wide academic trends.
OSPI has also adopted the student growth percentiles developed by Damian Betebenner as the state’s SGP scoring model. These model the growth trajectories of Star examinees to provide student SGP scores. The SGP score indicates how much a student grew relative to his or her academic peers, with a 100 point scale indicating a student’s growth was equal to or exceeded the growth of 75 percent of his or her academic peers.
SGP analyses are complex and involve several steps. Many of the errors that occur when running SGPs revert back to data preparation problems so there is often some back and forth between data preparation and analysis. However, once these initial preparation steps are completed, the actual SGP calculations are relatively straightforward.
The lower level SGP functions studentGrowthPercentiles and studentGrowthProjections require WIDE formatted data. Higher level SGP functions (wrappers for the lower level functions) require LONG formatted data. The exemplar data set, sgpData, is provided to model the format for using this type of data with these functions.
Researchers can access a significant amount of data from the SGP database by joining one of the Working Groups and contributing metadata for that Working Group. This is similar to how large community databases such as Genbank and EarthChem aggregate and make available essentially all of the data they contain. The primary difference between the SGP database and these other community databases is that the SGP database focuses on the research questions addressed by specific Working Groups. This approach allows the SGP database to access data that is unlikely to be contributed to a full community repository without the incentive of being part of an exciting research project.