Student Growth Percentiles (SGPs) place students’ achievement growth in context compared to the performance of academic peers statewide. SGPs are calculated using up to two years of historical MCAS test scores. The academic peer group used to establish Simon’s SGP is based upon his or her score history and the performance of all other students in the same grade and content area who have comparable test scores.
In order to create SGPs, students must have a valid MCAS test record with a scale score and achievement level recorded for each testing year. To calculate the SGP, a statistical procedure known as quantile regression is performed to place the student’s performance relative to academic peers on a normed scale.
SGPs are then used to report the average growth for a school/district/subgroup. In interpreting SGP reports, it is important to remember that percentile ranks are calculated anew each year and that differences in a student’s rank from one year to the next must be evaluated carefully.
This vignette describes the processes and assumptions that are involved in generating statewide SGP data. It is intended to help district teams understand the origins of SGPs, the processes and assumptions that are required to construct them, and how to interpret a student’s individual growth story.
A key part of SGP analysis is the creation of the sgptData_LONG data set. This file includes the sgpData_INSTRUCTOR_NUMBER lookup table and the anonymized, panel assessment data for 8 windows (3 windows annually) and three content areas.
Once sgptData_LONG is prepared, the higher level wrapper functions studentGrowthPercentiles and studentGrowthProjections can be used to produce a variety of analyses. These include SGPs for individuals, groups (school/district/subgroup), and the state aggregate.
The higher level wrapper functions require a long formatted data set with the following variables in the @Data slot: VALID_CASE, CONTENT_AREA, YEAR, ID, SCALE_SCORE, GRADE and ACHIEVEMENT_LEVEL. The data must also be normalized to a common scale by the function prepareSGP.
In most operational SGP analyses, all of the steps described above are conducted at once. The lower level functions are often nested within the higher level wrapper functions, making it easier to understand the source code associated with operational SGP analyses. This is especially true for the higher level functions, studentGrowthPercentiles and sgpProjections, which utilize the underlying wrapper function prepareSGP to perform their calculations. As an example of this approach, the vignette below uses studentGrowthPercentiles and printSGPs to demonstrate how the various steps of a nested SGP analysis are executed. For more information on SGP analyses, please refer to the sgpProjections and studentGrowthPercentiles guides or contact your OSPI Regional Data Coordinator. A version of this vignette is available in R Markdown here. If you have suggestions for improving this vignette or would like to collaborate with OSPI on other research and development related to student growth, please feel free to reach out. Thanks! We look forward to your contributions.