Data Sgp is a statistical software package for analyzing large scale, longitudinal education assessment data. The software uses quantile regression to estimate the conditional density of student achievement history and then creates percentile growth projections/trajectories based upon those projected densities. The resulting projections are used to identify students who need additional help to achieve proficiency in the current grade level and also provide guidance about how much progress the student will likely make in future years. This information is then used to adjust student learning goals in the Star portal.
In addition to the SGP analyses provided in this package, there are a number of additional functions and utilities that are useful for more general analysis of educational data. Some of these are available in the SGP toolkit on GitHub.
SGP combines a student’s performance on the state assessments with the student’s achievement history to identify the student’s relative standing compared to their academic peers nationwide. The calculation of SGP utilizes recent advances in the field of statistical methodology such as Betebenner’s well-known catch up/keep up growth projections.
In order to be assigned a teacher’s mSGP, a student must have taken the class for at least 60% of the course before the state test date. A teacher’s mSGP is only available for those teachers that teach English language arts or mathematics grades that have historical assessment data with comparable scores. Districts submit and certify course roster submission data to NJ SMART which is linked with the relevant SGP data. The final evaluation score that teachers receive is based on the mSGP of the class they teach.
A statewide median SGP score of 50 is calculated for each student by dividing the sum of a student’s growth percentages in the current year by the sum of a student’s growth in previous years. The median SGP for a particular school may differ from the state average in some cases due to slight misfit, the assignment of Highest Obtainable Scale Score (HOSS) students to an SGP of 99 or the use of different baseline methods for calculating SGPs.
The SGP package requires the free and open source software R to be installed on a computer before running any of its analyses. The package can be found on CRAN and is available for Window, OSX, and Linux operating systems. If you are unfamiliar with using R, it is recommended that you take some time to become familiar with its usage before running SGP analyses. In particular, the lower level functions, studentGrowthPercentiles and studentGrowthProjections require WIDE formatted data. The data set sgpData is an anonymized, panel dataset comprising 5 years of annual, vertically scaled, assessment data in the WIDE format. This exemplar data set models the format for data used with the lower level functions and should be used to verify that the analysis function are working correctly. For more detailed documentation on the usage of sgpData and wide format data sets in general, please refer to the SGP vignette.