Data SGP leverages longitudinal student assessment data to produce statistical growth plots (SGPs), which measure students’ relative progress compared to academic peers. A growth standard established using prior test scores and covariates is used as the measurement standard and then SGPs can be used to gauge whether students are meeting an agreed upon growth target (e.g. 75% of academic peers). Unfortunately, creating SGPs from students’ standardized test score histories involves complex calculations with large estimation errors that render such plots virtually unusable for measurement purposes.
Erroneous estimates are necessary in order to account for the fact that latent achievement traits cannot be directly observed with any level of precision (Akram, Erickson & Meyer 2013; Lockwood & Castellano 2015). SGP analyses typically utilize least squares regression modeling and Bayesian inference to estimate latent achievement trait models, and compare these estimates against growth standards established via teacher evaluation criteria and student covariates.
In theory, estimation errors can be minimized by comparing student SGPs against those from an identical baseline cohort of similarly-performing students in each year of measurement. The goal of such comparison is to establish an objective measurement of student progress that does not fluctuate over time and reduce uncertainty in making inferences about individuals or schools.
However, baseline-referenced SGPs require additional effort from students and data sources than their cohort-referenced counterparts, making them less accessible for use in educator evaluation systems. They need at least three years’ worth of stable assessment data in order to produce a model which can then be compared with scale scores from a prior year’s assessment data.
This practice can be problematic for several reasons. First, some teachers will be evaluated solely based on students’ current year growth without regard to whether they have yet met passing growth standards in class. Second, correlations between baseline SGPs and prior year scale scores likely won’t be exactly zero, potentially introducing substantial bias into interpretation of SGP results.
Baseline-referenced Student Growth Profiles may also be more vulnerable to spurious correlations with other variables, including teacher or school characteristics or design of baseline cohort. Furthermore, there is no guarantee that students in each of the five years necessary for baseline-referenced SGPs will remain under the same instructor throughout their educational experience.
Macomb and Clare-Gladwin ISDs have made their SGP data readily accessible in formats compatible with operational SGP analyses, making the data usable by districts. In order to effectively utilize it, districts require an accurate way of assigning students who have multiple instructors for one content area to one instructor for that content area – the sgpData_INSTRUCTOR_NUMBER field in sgpData table is an invaluable lookup function that allows districts to connect students with instructors through unique identifiers associated with their test records.