Teaching public health graduate students the statistics skills they need for their future careers has long been a challenge for educators. The standard approach used for many years has been giving students step-by-step instruction in statistical procedures. However, this method often leaves students without the conceptual knowledge needed to truly understand and effectively use statistics. More recently, educators have started adopting various concepts-based methods for teaching introductory statistics classes. These efforts have been largely successful, but such efforts have not been applied as extensively to more advanced statistics courses.
In a new study published in the journal Discover Education, Qi Zheng, PhD, professor at the Texas A&M University School of Public Health, explores the effectiveness of a concepts-based computational thinking approach for teaching longitudinal data analysis to public health graduate students. Zheng’s goal was to build a conceptual understanding in students with different mathematical backgrounds that would help them be better suited for using statistical tools as they move through their educations and careers.
Public health students often take advanced courses in statistics to better prepare for careers as researchers and practitioners. However, graduate students in public health have widely varying degrees of mathematical education. Some students will have a background in calculus, but many will have taken few mathematics courses since graduating high school. Teaching statistical principles couched in unintelligible mathematical jargon and complicated formulas is inappropriate in many cases. In contrast, using computational exercises and individualized assistance can help students work through underlying principles and continually build on their conceptual knowledge.
Longitudinal data analysis is among the most challenging subjects for public health students. The complex subject matter requires the instructor to find innovative ways to bring the underlying concepts within reach of most students who have limited prior mathematical training.
The effectiveness of Zheng’s approach rests upon many computational problems that he carefully designed to make abstract concepts tangible and comprehensible to students. Statistics is about understanding data by modeling data. To help students understand the meaning of a new longitudinal data model, Zheng gives students a small data set and asks them to rely on first principles to write down in a computer language the probability of observing the given data under the given new model. To catalyze the learning process further, Zheng gives students a similar worked example that he carefully devises to exploit the positive role of imitation in leaning. The worked example and its accompanying computational exercise are designed to encourage active learning and discourages mindless copying. Students then conduct computational experiments with their own code by changing input values to see how the see how the probability of observing the actual data varies with different input values. This hands-on process translates the elusive, yet all-important idea of the likelihood function into a tangible object on which students can enjoy experimenting. Students then compare their experimental results with output from a major statistical package. Students get immense intellectual satisfaction from this kind of hands-on exercise. Zheng’s approach makes it possible for students without strong mathematics backgrounds to understand the concepts involved.
This new approach has been very well received. The courses involved fewer students than introductory courses, and hence data is still limited. However, based on available course evaluation data, students were surprisingly receptive to the new approach. In a 2020 survey, students overwhelmingly agreed that the course was good as a whole. A 2021 survey shows that most students engaged in critical thinking and problem-solving skills. These results indicate high levels of student engagement, which points to the new method’s effectiveness.
Using computational exercises is an effective way to help students from a variety of mathematical backgrounds understand advanced statistical concepts. In addition, the process of writing code for exercises gives students ample opportunities to practice breaking large problems down and debugging errors, which helps students learn to think as a computer scientist thinks—a new competency for being an informed citizen of the 21st century.
Further work is expected to better fit computational thinking into statistics courses, but the results of this study show promise. Improved education techniques will help better teach public health graduate students as they prepare for research and practice.