Develop and Run SGP Analyses Requiring Knowledge of R Programming Language Creating and Running SGP analyses requires knowledge of R programming language. SGP is an analysis tool for student assessments in WIDE format (with some extensions to support LONG formatted data), suitable for Windows, Mac OSX or Linux and many resources are available to beginners beginning their journey with R programming language. Utilization of the SGP Package can be fairly straightforward following proper data preparation; however, errors that arise while running analyses typically relate back to either issues with data set preparation or data preparation steps themselves.
At the outset of an SGP process, data must be prepared by adding a header to an assessment data file. This step can be accomplished using the prepareSGP function from the SGP package; it takes WIDE format data with an additional special header that contains information needed to calculate student growth percentiles such as student identifier, year and grades associated with assessments in question; in addition, this data must contain scale scores for assessments taken between 2013-2018 in columns such as SS_2013_2014_2015 SS_2016 and SS_2017 columns of its data file.
Once data are ready for analysis, SGP functions may be run to calculate student growth percentiles and related outputs. The SGP package offers both lower level and higher level functions to help with analysis – such as studentGrowthPercentiles and studentGrowthProjections which require WIDE formatted data; summarizeSGP and plotSGP which require LONG formatted data with special header.
Student growth percentiles are calculated by comparing each student’s performance on this year’s MCAS test with that of academic peers who have taken one or more prior tests in the same content area. Quantile regression provides this comparison and allows us to establish their percentile rank.
At a state level, median student growth percentiles (SGPs) tend to fall near 50 because norms are often established based on only current-year test data analysis. However, average SGPs for schools, districts, or student subgroups can differ due to minor discrepancies in student growth percentiles on normed scales or other issues that arise in individual instances.
Student A and Student B both scored equally on this year’s MCAS test in one subject area; however, Student A made significant growth while Student B did not. This is likely due to Student A’s score on their previous test improving more quickly than most academic peers; by contrast, Student B had scores below the average for academic peers and thus experienced much less progress – hence why Student A’s SGP is much greater than that of Student B. This article is part of the Center for Assessment’s “Data Talk” series; please visit our website for all available articles in this series! This article forms part of our “Data Talk” series; please visit our website to read all available articles!