Teaching Philosophy

A new approach to teaching quantitative methods

I consider myself a part of a new generation of quantitative instructors that came of age after the invention of the personal computer and World Wide Web. We tend to take a different pedagogical approach that emphasizes learning by doing because we can. The traditional approach has its roots in the early 1980s, when data and the power and technologies needed to analyze it, were largely inaccessible. As a result it was forced to err on the side of abstraction. Over time this resulted in a tendency to teach the math that underlies quantitative research more than the actual practice itself.

Rather than begin with abstract maths, the new pedagogical approach begins by introducing students to data they can relate to. Data is basically a snapshot of reality, and is far more interesting and engaging to students than abstractions like the population mean. Moreover, most students are aware of the rise of data, and expect it to play a major role in their lives. Once they are engaged by data and what it can teach them, students are introduced to statistics as a tool for working with and interpreting it. Care is taken to ensure that new statistical techniques are introduced in a manner attuned to their skill level.

Introductory and intermediate courses and their progression

Over the last two years, I have had the privilege of teaching introductory and intermediate statistics at the undergraduate level at the University of Saskatchewan. These experiences have provided me with an opportunity to test my pedagogical approach, as well as familiarize myself with the capabilities of students at this level. On the whole, I have received very positive feedback from my students. I have had many tell me that they had believed statistics was not for them until they took my course. And even my students that have struggled have often voiced their appreciation for my approach and the respectful manner with which I engage them.

I have designed each of my undergraduate courses around five steps that follow the actual stages of quantitative research projects. Given the opportunity, I would retain the same structure for my graduate courses. The table below summarizes the common structure and skill progression I would use across all levels. Students begin each of my courses by identifying a research question, then reduce this question to variables and parameters appropriate to quantitative empirical study. This is a common stage in many sociology courses and helps to relate the practice of statistics to the wider practice of sociology.

Once students have a quantitative research question in mind, they are challenged to locate data that can be used to explore it. This is one of my favourite lessons because it provides me with an opportunity to introduce students to the wealth of data they have access to through initiatives like the Ontario Data Documentation, Extraction Service and Infrastructure (ODESI). Common reactions range from surprise to excitement. At the undergraduate level students select a Statistics Canada Public Use Microdata File (PUMF) to work with. At the graduate level, I will explore ways to grant students access to RDC resources, as has been done at other universities in Canada.

There is a common saying among quantitative researchers: “Garbage in, garbage out.” Data management is the practice of getting to know one’s data and transforming it into a format that can be analyzed to produce meaningful research results. Without data management skills, data analysis skills are of little use. In my experience, the data management step constitutes 80% or more of any quantitative research project. Nonetheless, data management is conspicuously absent from the traditional pedagogical approach. At every level, my students learn and practice data management techniques from opening and subsetting data at the lowest levels, to creating time-series constructions at advanced levels.

Only once data have been accessed and organized do I introduce the fourth step: statistical analysis. At the undergraduate level, I teach univariate statistics—which are powerful and useful in their own right—before progressing to bivariate techniques and OLS regression in the intermediate course. I deliberately move students along to regression as quickly as possible, because over the last two decades regression techniques have supplanted most others. For example, ordinal logistic regression has replaced Cramer’s V as the leading method for testing associations among ordinal variables. Ideally, students are prepared to learn advanced regression techniques when they begin graduate study.

My students at every level are taught using real data, and shown how to take into account survey design (e.g. weights) and missing data to produce publishable results. At the undergraduate level, students are asked to present their results in the form of a Descriptive Report that they would submit to an imaginary employer. At the graduate level, students will be expected to produce Analytical Descriptive Reports which are more suitable for academic audiences.

Finally, I teach my courses using leading statistical clients and course materials. At the undergraduate level, I teach the students using R, not only because it is available free online, but it is also quickly becoming the leading data analysis software used outside the academy. At the graduate level, I would use Stata, which is specially designed for the kinds of advanced multivariate regression techniques and complex survey data used by sociologists. I use OpenIntro Statistics as my course text for my undergraduate class primarily because of its adoption of the new pedagogical approach and its seamless integration with R. At higher levels my proposed course texts reflect my shift to teaching with Stata.