I've really been enjoying these two papers since watching a video presentation about the first one:

1. Shmueli, G., "To Explain or To Predict?", Statistical Science, vol. 25, issue 3, pp. 289-310, 2010.
- Video presentation: youtu.be/whD2sYFHW8c

2. Shmueli, G., and O. Koppius, "Predictive Analytics in Information Systems Research", MIS Quarterly, vol. 35, issue 3, pp. 553-572, 2011.

PDFs available here: galitshmueli.com/biblio/term/E

The presentation sensitized me to a few big things:

1. Research goals and theories can have either an explanatory/retrospectively descriptive nature or a predictive/forward-looking nature.
2. Depending on the type, the research can and likely should be conducted differently.

3. At this time, many fields, especially the social sciences, almost exclusively teach and use explanatory models and methods, even when their research goals are predictive in nature and would be better served by using predictive methods and models.
4. Predictive methods don't need to be used in isolation; they can be used to further the theory in a domain and its explanatory (causal) models. (See the second paper for *how*.)

Follow

Assuming that I'm understanding the second paper correctly, it even suggests that we can figure out which areas of theory (its explanatory/causal models) can likely be improved with greater or lesser amounts of additional effort (driving potentially greater research efficiency). This is by comparing the predictive accuracy of explanatory models to the predictive accuracy of predictive models, which are able to pull on potentially more sources of information.

Sign in to participate in the conversation
Scholar Social

Scholar Social is a microblogging platform for researchers, grad students, librarians, archivists, undergrads, academically inclined high schoolers, educators of all levels, journal editors, research assistants, professors, administrators—anyone involved in academia who is willing to engage with others respectfully.