Data SGP provides schools and students with comprehensive insight into their progress towards reaching academic goals, enabling educators to assess student growth relative to peers’ performance and identify areas of strength and weakness, informing the design of programs and interventions as well as program evaluation.
The SGP data includes student assessment scores from MCAS exams as well as local assessment results and educational tests, among other sources. To calculate an SGP score, a model compares current scores to those from the previous year and adjusts future ones to estimate what their expected performance would be; then percentile ranks are used to judge how well students are doing relative to their academic peers.
SGPs represent absolute rankings of students, making them easier to interpret and use for comparisons among them. It’s important to keep in mind that percentile rank changes are calculated annually; any differences in SGPs between years should be treated with care. For instance, a student with an MCAS assessment score of 60 would score equal or above 42% of his academic peers who have taken that test in previous years; these academic peers don’t need to share similar grade levels but must possess comparable MCAS score histories.
To conduct SGP analyses, one requires access to a computer running the free and open source R software environment, available on Windows, OSX and Linux systems and compilable on most other systems. Newcomers to R should first familiarize themselves with its usage before undertaking SGP analyses.
The SGP package provides classes, functions and data necessary for conducting SGP calculations. SGP utilizes quantile regression which uses conditional density of student achievements as input into future assessment outcomes to generate coefficient matrices that provide future assessment outcomes with their coefficient matrix values; these coefficient matrices then enable calculation of student growth percentiles as well as projections/trajectories.
Operational SGP analyses require higher-level wrapper functions (abcSGP and updateSGP) that make SGP calculations simpler. These functions combine lower level functions like studentGrowthPercentiles and studentGrowthProjections into one function call to reduce complexity associated with these calculations.
SGP analyses become straightforward if the proper steps of data preparation are followed, since almost all errors that arise in SGP analyses can be traced back to issues surrounding data preparation. By following this article’s guidelines and using tools from sgpData package for SGP data preparation you could save an immense amount of time in performing SGP analyses.
The sgpData data set contains five years of longitudinal education assessment data in WIDE format, providing anonymized student records with unique student identifiers as well as assessment scores from each of the past five years.