The Role of Theory in Data Analysis
What Does a Theory of Data Analysis Look Like?
Theory of data analysis binds together collective experience, summarizes what has worked and what hasn’t, compresses the past, and provides useful guidance for the uninitiated.
- The theory of data analysis is not a theory that instructs us what to do or tells us universal truths. Rather, it is a descriptive or constructive theory that guides analysts without tying their hands.
- A theory should speed up the training of an analyst by summarizing what is commonly done and what has been successful. It should reduce the amount of learning that must be done experimentally.
- The theory should serve the practice of data analysis and make data analysis better in order to justify its existence. This is in contrast to traditional statistical theory, whose existence could be justified by virtue of the fact that it allows us to discover truth about the natural world. In other words, a theory of data analysis would have relatively little intrinsic value.
- The theory should not be dependent on specific technologies or types of data. It should be reasonably applicable to a wide range of data analyses in many different subject areas.
- The theory should go far beyond the instruction of different methods of analysis, including aspects of data analysis that don’t strictly involve the data.