As data analysis tools become more powerful, the amount of information that can be gathered and analyzed has expanded exponentially. The term ‘big data” refers to datasets too large for traditional database systems and analysis techniques; by such standards sgp research does not use large datasets; nonetheless we are working hard on amassing an unprecedented amount of educational assessment data that may be analyzed using our custom relational database system.
Traditional student assessment reports provide you with details about a child’s performance on an individual test; growth reports provide more meaningful data by showing their rate of progress from year to year. Though there are multiple ways of interpreting growth data, one effective method would be comparing one student’s growth data against those of their peers.
Example: If Student A and Student B scored identically on this year’s MCAS test in one subject area, both students will earn an MCAS growth score of 70. If, however, one scored faster than another then one will achieve a higher growth percentile score indicating they outshone 70% of their academic peer group in growth percentile score.
At this point, it is crucial to realize that student growth scores reported in this context are calculated based on two years’ worth of experiences, so they may differ significantly between schools. This is particularly relevant to statewide average growth percentiles which represent all students across a state’s performance levels – educators should exercise extreme care when interpreting these averages, paying special attention to those students they know best.
The sgpdata package offers classes, functions and data for analyzing normative student growth percentiles (SGP) from longitudinal education assessment data. It follows procedures created by Damian W. betenbenner, Adam R. Van Iwaarden, Ben Domingue and Yi Shang at the Center for Assessment which utilize methods for estimating conditional densities from longitudinal education assessment data by inferring student conditional probabilities based on test score histories.
This data can then be used to calculate student growth percentiles and projections/trajectories. Furthermore, sgpData includes both wide-format (sgpData_WIDE) and long-format (sgpData_LONG) data sets that can be utilized for SGP analyses; long-format data is usually preferred as it allows for easier preparation and storage of the information, plus higher level functions like studentGrowthPercentiles and studentGrowthProjections are tailored specifically to work with long-format data sets.
Comparing different years’ SGP performance should be done with care, as percentile rankings change every year. Differences of less than 10 points should usually not be seen as being significant; this applies especially during years like 2024 when Covid-19 caused slower statewide growth compared with prior years.