Data SGP provides classes, functions and data used for performing student Growth Percentile (SGP) analyses. SGP analyses use large scale longitudinal education assessment data to estimate conditional density for individual student achievement histories and project/trace percentile growth projections/trajectories that highlight what additional performance gains would be necessary to meet future student achievement targets. Quantile regression calculations calculate these projections/trajectories using student historical assessment score data with associated coefficient matrices as inputs.
SGP analyses may be run on either WIDE or LONG formatted data sets, though for operational analyses we recommend LONG formatted data due to its ease of preparation and management steps associated with running SGP analyses. Furthermore, wrapper functions like abcSGP and prepareSGP make the transition between them much smoother.
The SGPdata package provides four sample data sets designed to be used with SGP analyses. The first, sgpData, specifies WIDE formatted data used with lower level SGP functions studentGrowthPercentiles and studentGrowthProjections; while sgpData_LONG and sgptData_LONG contain LONG formatted data used by higher level functions like abcSGP, prepareSGP, and analyzeSGP respectively. Finally, sgpData_INSTRUCTOR_NUMBER provides teacher-student lookup tables used by higher level SGP functions to produce teacher aggregates.
SGP stands out by using assessment score histories as the sole factor to compare students against academic peers; not demographics or program participation. Thus, SGP gives only an indication of how well the student has performed relative to academic peers over time and does not reflect overall performance across states and nations.
Thus, SGPs can be highly variable; during the Covid-19 pandemic for instance, median SGPs fell significantly for most students across most states. It’s important to understand that such fluctuations result from changes in performance data rather than any modification to SGP methodology itself.
Star Reports offer Window Specific SGPs when customized reports are selected in the Timeframe drop-down menu, enabling users to view specific SGPs from any prior or current school year.
As part of its analysis process, Star assigns each student to a cohort group based on academic peer history. A cohort group comprises all other students within their grade and subject who possess statistically similar assessment scores – this allows teachers to compare students’ performance within each year to previous year performances of academic peers in their cohort group.
To perform SGP analyses, a computer that is compatible with R, an open source statistical software available for Windows, OSX and Linux operating systems that is free to download via CRAN website is needed. You must also have access to Star Reports as well as students that you wish to conduct an SGP analysis on.