Although it is still not required by law that educators use Student Growth Percentiles (SGPs) for educator evaluations, many districts are using SGPs to help teachers identify ways to support student learning and growth. As part of this effort, the Educator Effectiveness Office has developed a number of helpful tools and resources for educators to use in their schools, including several video podcasts that explain how to create and use SGPs. These podcasts can be found on the Educator Effectiveness Office website.
The most important thing to remember about SGPs is that they are based on the comparison of students’ MCAS scaled score histories in order to determine their relative performance. This means that two students with the same MCAS scaled score history can have different SGPs because they likely have different academic peer groups. SGPs are calculated by comparing the student’s current MCAS scores to those of his/her academic peers from previous assessments and then assigning a percentile rank to this comparison.
As such, SGPs can be affected by the choice of academic peers, which is why we recommend that you choose the strongest academic peer group available in your data set for each individual student. This will help to ensure that the SGP estimate is as accurate as possible and will provide the most reliable results.
Another consideration when analyzing SGPs is that SGPs are typically reported at the state, district, and school/district/student group level rather than at the student-level. This means that average SGPs can fluctuate from year to year as the sample size of each group changes. However, if you are using SGPs to analyze the performance of your entire school or district, then these fluctuations should be minimal as the overall trend is indicative of your typical student growth.
On a statewide level, median SGPs are always approximately 50 because the norms for the scale are established using only student data from the current assessment year. However, when calculating SGPs for school/district or student group-level reporting, the median SGP may differ slightly from 50 due to slight misfit in the distribution of student test scores across the state.
When running SGP analyses, we strongly recommend that you use the sgptData_LONG data set which contains anonymized student assessment records in long format for each of the 8 windows (3 windows annually) of student test scoring that occur each year. The data set also includes the teacher-instructor lookup table sgpData_INSTRUCTOR_NUMBER which is used to associate a teacher with each test record. The higher level functions in the sgpdata package, such as studentGrowthPercentiles and studentGrowthProjections, require long formatted data sets. Managing this long formatted data is much simpler than using wide formatted data. Moreover, the higher level functions often assume the existence of state specific meta-data embedded in the sgpstateData data set. This is not a requirement, but it makes generating the analyses much more efficient.