I’ve taken a big dive into test analysis lately. Spreadsheets, formulas, correlations, and color-coding have become the norm after an exam to help and identify misconceptions with my students as a whole and on a class-by-class basis.
One of the sheets I created (based on work done by Andy Schwen a few years back) color codes student responses in red if they got the question wrong, like this:
I started wondering if there were significant patterns between classes looking at the scored matrix rather than statistical figures. I grabbed screenshots of each class, dropped the opacity of the image to ~35% and then layered them all to make this:
What made this doubly-interesting is that the darker red areas on the page do not always line up with the statistical figures. For instance, I needed to revisit 2, 4, 9, 11, and 12 based on the statistical analysis (biserial value, if you’re curious). They had discrepancies between students who got it right and did well on the test when compared to those who got it right and did poorly.
Looking at the picture, I also went back and revisited 8, 9, and 17 just because a lot of people missed them. I also looked at number 5 because very few people got it wrong.
Teaching is science and art. Even if the art is doing test analysis.