Data sgp is an R software package that provides classes, functions and data for calculating student growth percentiles and percentile growth projections/trajectories using large scale longitudinal education assessment data. It uses quantile regression to estimate conditional density from student assessment occurrence data and then translates those conditional densities into a matrix of derived coefficients that can be used to calculate both percentile growth estimates (for current students) and projections/trajectories for future achievement levels.
To use this data sgp, you’ll need a computer running the free and open source software program R. R is available for Windows, Mac OSX, and Linux. While it is possible to run SGP analyses with other statistical software programs, you will likely have more success if you familiarize yourself with the features of the R software package before diving into the SGP package.
For each student, SGPs are calculated based on the most recent assessment in a given testing window and at least one prior assessment from a different testing window. The most recent assessment is typically the spring or summer test for the previous academic year. If a student has taken more than one test in the same content area during a single testing window, additional rules are applied to account for student learning over time.
SGPs are calculated relative to the state average for a given test or subgroup of students. This means that a student that scores at the very top of their grade level will be compared to other students who scored at the very bottom of their grade level. This model helps ensure that the SGPs for students are a valid representation of their performance.
The sgpData data set is an anonymized panel data set containing 5 years of annual, vertically scaled, assessment data. This exemplar data set models the format of the data needed to run lower level SGP functions such as studentGrowthPercentiles and studentGrowthProjections. Higher level SGP functions, like the sgpData_INSTRUCTOR_NUMBER lookup table, require LONG formatted data.
SGP analyses are very straightforward once the data is properly prepared. In fact, any errors that may be encountered with SGP analyses usually revert back to issues related to the data preparation. For this reason, it is recommended that you always run your SGP analyses operationally on LONG formatted data. This will help to reduce the amount of work involved in managing your data set each year when updating analyses with new data. This will also make it easier to share data with others. Additionally, many of the higher level SGP functions are designed for use with long data sets and provide many preparation and storage advantages over wide formatted data. Moreover, these functions often take advantage of the embedded sgpData meta-data.