Student Growth Percentiles (SGP)

Gambling Jun 29, 2023

The sgp package provides functions for estimating, visualizing and interpreting student growth percentiles (SGP). SGP is a measure of the relative rate of a student’s achievement progress toward meeting the expectations of his or her academic peers. It is commonly used in educational assessment and reporting.

SGP can be derived from multiple sources of information. These include student performance reports from schools and teachers, teacher observation data, state assessments, classroom test scores, and prior achievement results. The sgp package also supports the use of student demographic and student categorization variables in addition to student achievement information. These additional variables help to identify specific student subgroups for which SGP estimates may be more sensitive and should be carefully considered when interpreting SGP estimates.

Sgp pools

The SGP package includes an exemplar WIDE data set, sgpData, and a LONG data set, sgptData_LONG. These data sets contain longitudinal student assessment information in WIDE format and the time dependent student covariates are spread across multiple columns of the data set for each student.

Using wide-format data like sgpData for SGP analyses is straightforward. See the SGP data analysis vignette for more detailed instructions. The sgpData_LONG data set contains the same student assessment information in a long format where each case/row represents a unique student and columns represent various time dependent covariates associated with the student over the course of several years. The first column, ID, provides the student unique identifier and the following 5 columns, GRADE_2013, GRADE_2014, GRADE_2015, GRADE_2016, and GRADE_2017 provide the grade level of the student’s assessment score in each of the 5 years.

As discussed in the SGP data analysis vignette, the within-subject, cross-year correlations of latent achievement attributes are likely to be more related to the student background characteristics than the student level covariates. This is expected since the student background characteristics are observed in each year and are used to match students with teachers, a process known as teacher sorting.

While this may explain some of the variation in within-subject, cross-year correlations for the student growth percentiles estimated in Figure 2, other factors could contribute to the observed differences. For example, the relationships between student background covariates and true SGPs could be due in part to the fact that different students are taught by teachers with varying levels of expertise, a phenomenon known as cohort effects. This may also partially explain why the correlations are smaller in Grade 7 and 8 compared to earlier grades. Further investigations are needed to determine the exact causes of the differences in correlations between student background covariates and true SGPs. However, these descriptive properties can be useful for evaluating the extent to which true SGPs are related to student background characteristics and identifying potential ways of improving the accuracy of SGP estimates.

By admin