Proposal for teaching data analytics
First published on LinkedIn on 14 February 2025. Lightly edited for republication, with direct links to the reading and example chapter.
I perceive a gap in the market for materials for learners of the fundamentals of data analytics.
There are excellent materials for the more advanced learner that cover machine learning or statistical analysis. But these materials generally skip all or most of the steps needed to create the data sets they use … they want to get on with the interesting stuff. But we all know that the hard part is often in constructing the data set in the first place. And surely many decisions in business and policy are made on the basis of a compelling plot rather than a p-value or AUC statistic (am I wrong?).
There are also excellent resources for those looking to learn the nuts and bolts of data science. But these often require the user to supply the motivation. Not every student is excited by the flipper length of penguins, cute as those birds might be.
Finally, it seems that many courses in business and economics miss the opportunity to learn key concepts by messing around with data.
Given the above, I have identified a number of desiderata for a learning resource. For concreteness, I assume that it’s a book I’m looking for:
- Minimal assumed knowledge. No statistics or economics beyond what the reader might’ve learned in high school.
- Real data. Real data allows motivated students to go further and ask deeper questions.
- Core concepts from business and economics. Starting almost from scratch, readers should get an understanding of concepts such as GDP, inflation, stock returns, profitability, and more.
- Minimal use of statistics. No machine learning, perhaps not even regressions. One can go far with tables of descriptive stats and plots.
- Extensive use of data visualisation. A graduate of a good business program in 2025 should be able to transform data into insightful plots.
- Complete code. A reader should be able to generate all plots and analyses in the book. No hidden steps!
- Coverage of core skills in data science. A studious reader of the book should end up with a level of proficiency close to that provided by R for Data Science. The reader should be able to go to work with “real” data sets (in 2025, a laptop can—with the right tools—handle data sets over 100GB surprisingly easily).
I haven’t quite found this book (open to suggestions). Meanwhile, I have started to cobble together some materials of my own. Retail sales is a (very rough!) chapter along these lines. (I see this coming somewhere around the middle of a hypothetical book, so I can assume that the reader will have read something like “Describing data” from Empirical Research in Accounting: Tools and Methods.)
I recognise that this is a bit tedious in places. Time may give me ideas to bring some sections to life. But it covers a lot of ground and I think the closing example (is there an effect of Covid lockdowns on liquor consumption?) illustrates what you can get to with just a few lines of code and a plot.