Unlike some ‘big data’ analyses that use massive amounts of information and complex applications to manage the analysis, SGP research is focused on small to medium size datasets. While it is still quite a significant effort to assemble this kind of information it is much easier to manage than the billions of observations and analytical results that might be produced by analyzing global Facebook interactions.
This vignette demonstrates how the prepareSGP function in the SGP package helps simplify this process and provides the foundation for more sophisticated analysis. Once the data is prepared it is then possible to generate growth achievement charts using the plotSGP function.
SGP stands for Student Growth Percentile and is a measure of relative achievement in which students are compared to students who have similar prior performance. Specifically, an SGP is the percentile rank of a student’s current achievement on MCAS compared to other students who have similar prior test scores. This is a familiar and interpretable measure of relative progress even when standardized test scores are not vertically or interval scaled.
When comparing student SGPs, the choice to use means or medians is an important decision. The median SGP is computed by selecting the middle value of the observed scale score distribution (i.e. the point that is 50% above and below the average). While this approach is common in SGP literature, it can be a misleading indicator of school performance. This is because the distribution of SGPs for students in a given school often does not follow a normal bell-shaped curve, and instead is clustered around an average SGP.
This is due to estimation error in the estimated SGPs, which are a function of the previous test score and teacher fixed effects. These errors can be a source of bias in the interpretation of aggregated SGPs as an indicator of teacher effectiveness. This bias can be avoided by modeling the student SGPs with a value-added model that regresses the estimated teacher effects on the students’ prior and current test scores.
Other useful information that can be retrieved from the sgpData data set includes a sgpData_INSTRUCTOR_NUMBER table, an anonymized lookup table that allows teachers to associate their instructor number with student test records. It is also possible to access sgpData_CONTENTS tables that describe the amount of material covered by a given subject for each year, which can help educators identify areas where students may be lacking knowledge and inform curriculum decisions.