The sgpData vignette provides comprehensive documentation on using wide format data with the SGP package. It’s worth reading if you plan on running SGP analyses operationally year after year or you are interested in learning more about the SGP package and its lower level functions. For most SGP analyses, you will probably be better off formatting your data in LONG format and using the higher level wrapper functions like studentGrowthPercentiles and studentGrowthProjections. LONG formats have many preparation and storage benefits over WIDE formats.
The SGP package has been designed with two primary uses in mind: creating a data frame and calculating student growth percentiles. The sgpData vignette explains how to use these tools together for a complete workflow.
A student’s growth on a state assessment is measured as a relative percentile. The percentile represents the point at which a student’s score is compared to the scores of students with similar prior performance history. For example, if a student’s SGP is 85, she has scored higher than about 85% of students with similar MCAS history.
To calculate a student’s growth percentile, we take the student’s current MCAS scale score and divide it by their prior MCAS score. The result is a number between 1 and 99, with higher numbers representing greater relative growth. In this way, a student with a low initial score can show high growth and a student with a high score can have relatively little progress.
One could calculate a student’s growth using only their current and prior test scores, but this ignores important differences between assessments such as the specific grade-level and the degree of difficulty associated with making each assessment type’s scale score change. It also does not take into account the fact that a student’s true SGP may be related to their background characteristics.
Recent research has shown that SGPs estimated from standardized tests have large estimation errors (Akram, Erickson, & Meyer, 2013; Lockwood & Castellano, 2015). This makes the estimated SGP noisy measures of their corresponding latent achievement traits. However, it is possible to construct a more accurate measure of the true SGP by ranking students against other students with similar previous achievements rather than comparing them to all students. This method has several advantages over the previous approaches described above. It is more objective, fair, and valid for evaluating student growth and educator effectiveness. These benefits have contributed to the growing popularity of SGPs in the United States.