Data Minding (CI15879)

General information

Type:	OP	Curs:	1	Period:	S semester
ECTS Credits:	3 ECTS

Teaching Staff:

Group	Teacher	Department	Language
	Uri Simonsohn	Operaciones, Innovación y Data Sciences	ENG

This course focuses less on technical aspects of data analysis and more on conceptual ones. It focuses on how we should be thinking about data before and after running the statistical tools students learn in other courses. The course content is primarily based on the following three sources of information:
(1) Psychological research on people's incorrect intuitions about data and randomness,
(2) Practical tools empirical researchers have developed, mostly in medicine, economics, and political science, to make firmer causal inferences from data. These tools are useful also for everyday decision making in organizations but solve problems that are typically assumed away in traditional statistical textbooks and courses, and
(3) Foundational statistical concepts which are also useful for everyday decision making in organizations but which are usually too abstractly covered to be implemented in everyday practice (e.g., statistical power, practical vs statistical significance, concluding something does not exist vs failing to conclude it does, etc.).

Improve students' ability to:
1) Figure out what is the right way to ask a given question from data, and
2) Understanding what question a given analysis is actually answering.
For example, say a company rolls out a promotion to increase sales. Comparing sales before and after the promotion does not tell us if it was successful.
This course helps students better understand what question the before after comparison does answer, and how to adequately asses if the promotion worked.

1. Content 1) Statistical bootstrapping (to conceptually understand black-box solutions in classical stats and to carry out analyses for which formulas do not exist) 2) How to evaluate if an intervention was successful? 3) The challenge specifying the question after we see the data (that vs like-that and mutiple comparisons) 4) Experiments as information 5) Diagnosing and correcting for self-selection 6) Disentangling skill from luck in performance 7) Forensic statistics (how to detect errors, selective reporting, and fraud) 8) Learning from errors and close-calls 9) How the choice of metrics (the ways we measure something) influences what we conclude 10) How much data is needed to answer different questions 11) How to analyze data without making assumptions about normality or linearity (i.e., in the real world).

Lecture discussion and hands-on analysis of data in class. We will often contrast how students solved a (practical, real-world) problem before coming to class with how the content covered suggests we cover it. All required analysis can be performed in Excel and will be taught in Excel. If there is student interest, and only for students who are themselves interested, we will also learn how to conduct those same analysis in R (an open source, free, and increasingly popular statistical program).

There is no textbook for this course. For most classes background readings will be recommended rather than required.
They will not cover the same material that's covered in class.

Group	Teacher	Department
	Uri Simonsohn	Operaciones, Innovación y Data Sciences