Data sgp is the collective of aggregated student performance measures collected over time that teachers and administrators use to identify areas for improvement, inform classroom practices, evaluate educators, and support broader research initiatives. Often referred to as “growth data”, the most important part of data sgp is growth percentiles – measures of a student’s relative progress compared to peers with similar MCAS score histories.
Creating growth percentiles from standardized assessment data requires generating complex statistical models that incorporate prior test scores and covariates, which can lead to large estimation errors and render these estimates virtually unusable for measurement purposes. This is why the data sgp method is increasingly being used in place of traditional growth plots to determine whether students have met an agreed upon “growth standard.”
An SGP provides a quick and easy way to measure the relative progress a student has made on MCAS by ranking them against the progress of other students with similar score histories. SGPs are reported on a scale of 1 to 99, with higher numbers indicating greater relative growth. For example, a student with an SGP of 75 has scored better on a recent subject-matter test than 75% of students with comparable MCAS score histories.
A growing number of states are using SGPs to assess educator effectiveness and report on student achievement. However, while SGPs can be a useful tool for educators and policymakers, they are not without limitations. Specifically, SGPs can be vulnerable to spurious correlations between teacher-level characteristics and student growth (e.g., motivation, skills, or family circumstances).
Additionally, since SGPs are based on the comparison of individual student assessments with a statistically constructed baseline cohort they can be biased by design factors such as the composition and length of the initial assessment battery. Finally, SGPs can also be affected by contextual effects – that is, the fact that a student may have more or less growth as a result of the particular classroom in which they are taught.
To address these limitations the sgp data package offers an alternative to traditional growth plot calculations that uses machine learning methods to generate SGPs from state assessment data. The package utilizes the open source R software environment and includes a series of functions that facilitate the calculation of student growth percentiles from longitudinal data sets. The package is available for download on the sgp data github page. It is important to note that while most analyses that the SGP package supports can be conducted with WIDE format data, for operational analyses year after year it is generally preferable to use the LONG data format which allows for easier preparation and storage. Please consult the SGP package documentation for more detailed instructions on the difference between these two formats.