In short, a student growth percentile (SGP) measures how a student is performing in comparison to students with similar prior test scores (their academic peers). It is important for teachers to understand this measure as it can have an impact on their evaluation.

In Washington, a student’s SGP is calculated using assessment data beginning in 2005-06 for the subject and grade level of their MCAS testing. A student’s academic peers are all other students in the state in that same grade and assessment subject who have had statistically similar test score paths, regardless of their demographic characteristics or program participation. The SGP provides a snapshot of how the student performed relative to those academic peers, and it can tell us whether a student is progressing toward meeting or exceeding state standards in that area.

The SGP model uses linear and nonlinear mixed effects models to estimate student performance and its associated predictors. Students in the same cohort are grouped into groups, called clusters, based on their MCAS scale score histories and then compared to one another. Those with the highest scale scores are assigned to the top cluster, and those with the lowest scale scores are placed into the bottom cluster.

Each cluster is then modeled using the same model. The model produces a predicted value for each student in the cluster and for each of their predictors. For each predictor, the model also calculates a confidence interval which provides an estimate of the range of possible values for that variable.

SGP models are designed to be as accurate as possible. However, due to the computational complexity of Gaussian Process regression models – requiring O(N3) time and memory for a large N – it is impractical to use them on very large datasets. Approximation methods, such as sparse GPs and variational inference, have been developed to help overcome this limitation.

The sgptData_LONG data set is an anonymized, panel data set comprising 8 windows (3 windows annually) of assessment data in LONG format for 3 content areas. The data set contains 7 required variables – VALID_CASE, CONTENT_AREA, YEAR, ID, SCALE_SCORE, GRADE and ACHIEVEMENT_LEVEL (on required if running student growth projections).

The purpose of the sgptData_LONG is to demonstrate how this data can be used to create student growth percentiles and student growth projections. This data set models the format of data that is submitted to NJ SMART and the mSGP function.

In order for a teacher to receive an mSGP for their teaching in 2023-2024, they must have valid SGP data for all of the following: