We compare student performance in two sessions of a large first-year university physics course, one with a normal 12-week term and the other with a compressed 6-week term. Student performance is measured by the normalized gain on the Force Concept Inventory. We find that the gains for the regular-format course are better than the gains for the compressed-format course, and while the differences in gains are small they are statistically significant. Not accounted for are the differences in effectiveness of the different instructors in the two versions of the course.
If the uncertainties were based on the standard error of the mean, twice the uncertainty's value would be a 95% confidence interval, but for the median, as discussed in Ref. 11, the interpretation is more indirect.
This way of characterizing the effect size is not without controversy and can be overused. Romano, Kromrey, Coraggio, and Showronek go even further and recommend setting thresholds of “negligible,” “small,” “medium,” and “large” otherwise, in “Exploring methods for evaluating group differences on the NSSE and other surveys: Are the t-test and Cohen's d indices the most appropriate choices?,” presented at the annual meeting of the Southern Association for Institutional Research, October, 14–17, 2006, Arlington, Virginia. http://www.coedu.usf.edu/main/departments/me/documents/MethodsforEvaluatingGroup.pdf (retrieved September 29, 2014).
The lower value is the 95% CI infimum and the upper value is the supremum.