Data sgp is a data format used by the SGP package to hold information about students and their achievement. The format can be useful for analyzing data with a wide range of methods. It contains a unique student identifier, scale scores associated with students in each year of testing, and covariate values. It can be used in the same way as WIDE data formats and is supported by many sgp packages.
The data format is available from the sgpData vignette and the SGP package documentation, and it is recommended that users consult these sources for a fuller description of how to use this data type. It is generally straightforward to use the sgpData for analysis of SGP data, but please ensure that you have read and understand the SGP data analysis vignette before attempting to use this data format.
True SGPs are Correlated with Student Background Characteristics
The distribution of true SGPs relates to the underlying characteristics of students, which are defined as latent achievement attributes. The distributional properties of these attributes are described in the following section, where we also explain how to interpret relationships between true SGPs and student covariates at the individual level.
These relationships between latent traits and student covariates create a problem for interpreting aggregate SGPs as teacher performance indicators, even when the aggregate SGPs are sorted to teachers of equal effectiveness but who teach different types of students. They would cause teachers of higher quality to receive systematically lower aggregate SGPs than less effective teachers. This effect could be caused by a number of mechanisms.
One mechanism is that more effective teachers teach a larger proportion of students with certain background characteristics than less effective teachers do. This would lead to a greater percentage of the variance in aggregate SGPs being due to these relationships, as opposed to other mechanisms. It may also be that more effective teachers are more likely to be assigned to schools and classrooms where their students have particular background characteristics.
Another possible mechanism is that teachers who are more effective tend to have more of their students in the same background group, such as a socioeconomically disadvantaged group. The higher the percentage of students in a given group, the higher the average group mean difference in true SGPs. The average difference in the group means is higher for math than ELA.
This could have a large effect on the ability of teachers to estimate SGPs, but the extent to which this is the case depends on the type of relationship between true SGPs and student covariates. This type of relationship, which is more common in the higher grades, is not as evident for lower grades.
In general, the distributional properties of true SGPs for students are correlated across math and ELA, and are related to their student background characteristics. They are also correlated with a range of time-varying covariates, such as gender, race/ethnicity, home language, and so on.
The relationships between student background characteristics and true SGPs can be used to improve the accuracy of estimated SGPs at the individual level, as the relationship between a student’s real SGP and its covariates is often correlated with its measured SGP. This can be useful in identifying students who are at high risk of failing to meet the SGP benchmark or who are at low risk of reaching the SGP threshold.