Admin at a university would like to know what language classroom activities have the greatest impact on the tests we use for benchmarking/completion of each level in our language program. The problem is that each version of the level test has a floor and ceiling, so there is a limited range of scores, with some students getting no score, or reaching the maximum score. I think this is a standard statistical problem in many fields, so I guess we could learn from other fields about options that are tried and true for designing studies.
Thanks for your input on this!
Dear Jeff,
Thanks so much for your insightful response. Our students take these benchmark tests twice every semester, so it would be possible to consider replacing scores from those who score too low on the current range version with those from the previous range version that was several weeks prior. That might be one way to extend the range. I hadn't thought of that. Also, for those that go through the ceiling, we could also take their next test score on the following range level, with a caveat that we use test scores from one academic year or something. I also spoke with another testing colleague who said that students below the floor could perhaps be offered to take the test in the lower/previous course-level range, which is an interesting idea... I will also try the z-score idea and see how that would work. I am not familiar with Winsteps but will look into this --- thanks ever so much for your input... Paul
- Badges
- Science topic
- Similar topics
- Mathematics
- Statistics
- Applied Statistics
- Survey Design
- Questionnaire