Data SGP and Statistical Growth Plots

Data SGP (Student Growth Percentiles) are collected over time by teachers and administrators for use to inform instruction, evaluate students/teachers/schools/districts, support educator evaluation systems, conduct research, support educator evaluation systems and conduct research. Data sgp includes both individual-level measures like test scores and growth percentiles as well as aggregate measures at school/district levels like student achievement trends and class size trends.

Statistical Growth Plots (SGPs) use longitudinal student assessment data to create statistical charts that measure relative student progress based on prior test scores, comparable peers or content areas. Though SGPs provide educators and parents with a useful metric for tracking progress against similar academic peers with similar test histories, creating one can be challenging when students’ past test history includes multiple testing windows or content areas. Furthermore, creating SGPs involves calculations susceptible to spurious correlations between prior test score data and school/teacher characteristics as well as cohort design variables – further complicating its construction.

The R sgpdata package provides tools for conducting SGP analyses using its open source programming language R. To use it effectively requires at least basic familiarity with R, as running SGP analyses requires understanding how to prepare data for use with studentGrowthPercentiles and studentGrowthProjections functions and using higher level wrapper functions such as abcSGP and updateSGP functions that simplify source code for conducting operational SGP analyses.

As part of preparing data for SGP analyses, it is essential to keep in mind that the lower level functions require WIDE formatted data. The sgpdata package offers two exemplar LONG data sets (sgpData_LONG and INSTRUCTOR_STUDENT lookup table; see also: INSTRUCTOR-STUDENT lookup table (sgpData_INSTRUCTOR_NUMBER), to assist with this preparation process and use this lookup table with instructors/students lookup table when creating SGP analyses. In addition, the vignette provides further assistance when using these sets when carrying out analyses using these sets for implementation of analyses using these data sets when conducting analyses using SGP analyses.

SGP analyses require extensive computing power. Therefore, optimizing data structure and reducing the number of variables utilized is vital. One common optimization method involves preprocessing the data by eliminating redundant elements like test scores, school and teacher information to reduce computational requirements significantly while increasing accuracy of estimates from SGP estimation models.

Preprocessing can reduce both computational cost and reliability by eliminating covariances between test score predictors and student-level covariates that affect estimates from SGP methods, helping to significantly decrease estimation error when comparing latent achievement trait models estimated with this approach to standardised test score histories used as growth standards; it is particularly helpful when making comparisons across groups of students/teachers/schools/districts; it plays a pivotal role in strengthening both consistency and validity of SGP estimates.


July 2024


Recent Posts