Everyone knows that American colleges and universities are doing a poor job of teaching most disciplines. History, sociology, anthropology, English, and other subjects have been infiltrated by faculty who convey opinions rather than impart knowledge.
What about a hard, technical subject like data analysis? While it seems (so far) to have escaped politicization, it isn’t being taught in an optimal way. So argues Steven Zhou, a Ph.D. candidate, in today’s Martin Center article.
Zhou explains that data analysis has advanced rapidly in recent years, and programs in it cover the advances. The problem, he argues, is that students are left deficient in some basic skills.
Zhou writes, “By over-emphasizing these advanced analytics methods, many programs are losing sight of the importance of ‘simpler’ methods and skills that are actually more important and relevant in future careers. My colleague and I collected an informal survey of about 100 alumni working in non-academic jobs, asking them what statistical methods they used most frequently at work. The most frequently used methods were simple correlation (62 percent used ‘a lot’), data visualization (55 percent), and regression (49 percent); the advanced methods taught in most curricula were only used ‘a little’ or ‘not at all.’ In fact, the most frequently used software was Tableau, which is a platform for data visualization—this reflects the growing trend of data visualization as the key skill in analytics jobs.”
The big failing, Zhou contends, is that students aren’t trained in data-visualization techniques: “the repercussions of poorly conducted data visualization and the inability to explain statistical findings are potentially far more damaging . . . for the vast majority of the population, advanced methods are much less valuable than the ability to communicate and visualize data. It does no good for students to be able to perform a ‘latent class analysis’ if they are unable to explain the method, demonstrate why the findings are important, and visualize the results to people who have no idea what a latent class analysis is.”
I have heard it said about other technical academic programs, such as computer science, that they tend to center on the things that faculty members like to teach more than the things the students need to know. Apparently this is true of data analysis too. Deans ought to pay attention.