A student growth percentile (SGP) describes a student’s relative progress on MCAS by comparing their current score with the scores of students who had similar prior test performance histories (their academic peers). SGP scores are measured on a scale from 1-99, higher numbers indicate greater relative growth. Teachers and administrators can use SGP data to determine whether a student has risen above, below or as expected relative to their academic peers.
SGP analyses are designed to provide information on student achievement in a format that is both informative and user friendly. However, the analysis requires some familiarity with the software environment R and a willingness to invest some time in proper data preparation. Any errors that occur when running SGP analyses usually revert back to the data preparation process and are best avoided by following the recommended procedure.
The SGP package is free and open source, available for Windows, OSX and Linux. To run SGP analyses you will need a computer with at least one of these operating systems and a copy of the R software. If you are not familiar with R, we recommend taking some time to get familiar with it before diving into SGP analyses as there is a good chance that the majority of your time will be spent working in the application.
In addition to individual student SGPs, SGP data can be aggregated to summarize the progress of subgroups, schools and districts. Historically, median SGPs have been the primary summary statistic used to represent this aggregated data. However, the Department is now switching to mean SGPs as the default choice for these aggregated analyses. This article explains the rationale for this change and the advantages of using means rather than medians.
Statistical growth plots (SGP) are a valuable tool in assessing the progress of students, but can be subject to large estimation errors that can distort the true nature of a student’s achievement. In order to reduce these estimation errors, SGPs are often combined with latent trait estimates to construct what is known as a composite score. While this method can reduce estimation error, it does introduce noise into the measure of student achievement and may not be appropriate for all uses.
SGP analyses rely on a number of variables including the student’s assessment history, grade level, content area and the prior year’s test score. In order to accurately compare the results of SGP analyses it is important that all of these variables be identical across the entire dataset. As a result, when using the sgptdata_LONG data set, the following variables must be included for each student: VALID_CASE, CONTENT_AREA, YEAR, STUDENT_ID and SCALE_SCORE. The sgptdata_WIDE data set also includes the demographic variables but is not required for SGP analyses.