4 ms·
Based on the graphs given, it looks like the low variance may simply be a result of having the x-axis be percentile. Consider Round 2, where sigma=1.2%, mu=98%
by yummyfajitas 16y ago
Based on the graphs given, it looks like the low variance may simply be a result of having the x-axis be percentile.
Consider Round 2, where sigma=1.2%, mu=98%. Suppose hypothetically that 90% of students are clustered below 100 (an absolute measure of performance), 96.8% of students are below 150 pts, and 99.2% of students are below 300 pts.
In this case, the absolute variation is huge (300 vs 150). But because only 2.4% of students score between 150 and 300 pts, on a percentile graph, it looks like sigma has been reduced.
I think it's unlikely that this has occurred, but the graphs given don't preclude it.