Tuesday, January 3, 2017

2016 Student Growth Percentiles

The Georgia Department of Education (GADOE) calculates Student Growth Percentiles (SGP) for students who take the Milestones exams. SGPs indicate how much a student's achievement grew compared to academically similar peers. To calculate a 2016 SGP score, a student is grouped with other students with similar 2015 and 2014 scores, and then given a percentile rank based on how their 2016 score compares.

GADOE then uses SGP data to find the percentage of students making typical or high growth by school and district. "Typical or high growth" is defined as an SGP of at least 35. The state average percentage of students making typical or high growth is 65% (100 - 35).

The visual below shows the percentage of students making typical or high growth by grade and subject, using SGP results from GADOE. Use the filter on the left to choose a different school or district, or the tabs at the top to switch to high school or district-level views.


In this example, we can see that students at Inman Middle School show progress at rates above the state average in most grades and subjects. Seventh grade social studies was a relative weakness in 2014, but has since recovered to near the state average. Overall, math is a relative strength, where 80% of students are achieving typical or high growth.

This visual is helpful for schools or districts to use root-cause analysis. Working with other school or district leaders, identify which grade/subject combinations are high or low-performing, or have seen significant change. Then discuss the causes of those notable data points, and how that information can be used for future improvement.

The typical or high growth calculation is also used to determine progress points on the CCRPI. However, the CCRPI only includes "full academic year" students, while the SGP data files include all tested students. The results used for CCRPI progress points might vary from the SGP files by a few points but they show the same trends.

Confidence intervals1 are displayed as blue bands. Similar to the margin of error in a poll, the confidence interval tells us the range of possible values, but the middle value is most likely. Schools or districts with few students will have wide confidence intervals; large schools and districts have more precise growth estimates so their confidence intervals will be smaller.

Although progress data is very helpful for understanding school performance it does not give a complete picture. It is also important to consider overall achievement. This is especially true in elementary school, where progress is only measured in two of the six grade levels. A helpful compliment to the progress data is the Milestones versus Challenge Index graph below. This compares achievement for schools with similar populations, and better accounts for learning that occurred before 4th grade.


For more details on this visual, see our post from last year. This year mobility has been added to the challenge index because it adds additional predictive power2. Mouse over the details mark for more information on the new challenge index calculation.

(1) Confidence intervals are calculated at the 95% level using the standard error of the sample mean. This is only one of two sources of uncertainty. Each student SGP score is also an estimate and has associated uncertainty, but is not available to us. Hence, the confidence intervals are an underestimate of the total uncertainty. We compensate for this by using a relatively large confidence interval- 95%.
(2) Mobility improved R-square from 93.3 to 94.4 for elementary schools. Charter schools are not included because of different mobility patterns due to grade bands and school zones. Similar graphs that include charters for 2016 can be found at the bottom of this article.