I knew I was in trouble teaching statistics when I always thought I was doing such a great job, but the students were totally lost the last four weeks of the semester! It didn’t matter how many times I used active learning in my classroom or the how many great lectures I thought I presented, they just did not get the concept of a hypothesis test. [pullquote] it is wonderful to present the ideas of how to think about research questions, the proper way to write hypotheses, the meaning of a p-value, and the meaning of statistically significant within the first four days of a semester.[/pullquote] The early part of the semester with data description, probability, and experimental design had lulled them into a false sense of security.
When I heard about the simulation-based, or randomization-based, approach to be presented in a 2013 MAA summer workshop, I decided to chuck everything overboard and resupply with a new method. Several other professors attending the workshop told me I should not throw out everything I had used for 15 years but instead should integrate a few of the new ideas throughout the semester. I knew that was not enough of a change. How could I use the randomization- method if I didn’t start on Day 1? I knew I was in – for better or worse.
The past year of teaching with this new method has been a challenge as most new methods are, but I am seeing very positive results. It is wonderful not to have to present data description for 2-3 weeks, both graphically and numerically, material they had learned in elementary-high school. How boring was that? Instead, it is wonderful to present the ideas of how to think about research questions, the proper way to write hypotheses, the meaning of a p-value, and the meaning of statistically significant within the first four days of a semester.
Does this mean the students understand all of these terms in four days? Oh no, it takes a whole semester for these ideas to become ingrained, but they are not in shock any more about being presented with new procedures at the end of the semester when they are already stressed with other classes. Rather than trying to teach the z-test and t-test for one mean in a semester, I had the ability to teach the t-test for both one and two means, paired t-test, one/two proportion tests, Chi-square test, ANOVA, and regression tests for the slope. Wow – the students are eating a real filet mignon instead of ground meat! Their grades are actually better in this course as well. Most can earn an A or B, with a few Cs, but rarely do I have anyone earning less than that. The material is constantly repeated in their brains for 14 weeks – what wonderful reinforcement. There are no surprises at the end to scare them to death.
The end of the semester class project was reaffirming that the randomization–based method was working. In the past I always had students decide on their topic, collect the data, run the analysis, and present their project to the class after completing all of the course work. The past two semesters, however, the projects have been much more exciting because the students can work on it as we move through the book, and the presentations are outstanding. The wide variety of hypothesis tests they used was fun to see and watching them decide on the proper test to use was interesting as well. They realized that the project was not that much more difficult than a standard problem we had worked in class, and this gave them much confidence.
It is interesting that what used to be a freshman analytic reasoning core class for all Butler University students is finding its place for those who need to use research for topics in their major coursework or for students who need a statistics class for future graduate programs. There were easily 14 of 24 students in one class in the 2014 spring semester who planned to go to graduate school and needed a basic knowledge of statistics for their future academic work.
Nothing is perfect, and I have to admit this method is not easy to teach. Each teacher must find the best way to get the students to work through the material, but I would definitely prefer not to teach a traditional statistics class after this experience. The thought of doing that makes me realize how lucky I was to find such a great method taught by such wonderful people who could help me see the light! I hope my students get half as much out of the class as I do.