MYERS, Marcella J.; College of Saint Catherine: The Confidence to Use Confidence Intervals:Moving Students toward More Sophisticated Analyses
As the result of much trial and error in the undergraduate classroom, I have become convinced that: 1) students require objective rules (statistics) to properly interpret their data, no matter how simple the experiment, 2) students learn statistics most effectively while doing science, and 3) students should be introduced to statistical concepts as early as possible, as it takes repeated exposures for real understanding to take place. In the past, I have found that introductory biology students are quite capable of using standard errors (essentially hypothesis tests) to judge whether two means are significantly different. The challenge with these students has been to help them meaningfully interpret the statistical outcomes of their experiments in the context of the biology involved; students tend to conflate statistical significance with biological significance. Recently, my colleagues and I have begun teaching our students how to use confidence intervals to analyze their experimental results. Confidence intervals give upper and lower bounds for the likely size of any true treatment effect, rather than just determining whether the treatment effect is significant. This approach seems to help students look beyond the dichotomy of significant vs. non-significant treatment effects (which can also be determined using confidence intervals) and to instead focus on the range of possible effect sizes. Thus, for both significant and non-significant differences, students are required to think about and discuss the size of treatment effect that would be biologically meaningful in the system under consideration. In my talk, I will discuss these issues more fully and show the presentation on confidence intervals my colleagues and I give early in the semester in our introductory biology labs.