Tracking the right clues with exploratory data analysis

Abstract

This alternative to traditional statistical analyses is guided by clues uncovered in the course of an investigation. Exploratory data analysis can be likened to the examination of a crime scene. Just as a good detective approaches an investigation with an open mind and as few preconceived notions as possible, so, too, should data analysis be guided by clues uncovered in the course of the investigation. In contrast, the traditional statistical approach may fail to recognize unexpected behavior or anomalous data if the method of analysis turns out inconsistent with the data. In computer-based exploratory data analysis (EDA), linked tools serve to analyze the data in an interactive, free-flowing process of discovery, without constraint to a preconceived path. Open-ended exploration along these lines is often more fruitful than simply pouring data into a fixed analysis path designed to spit out "the answer" or churn out statistical summaries. The article discusses the essential features of exploratory data analysis. The technique is illustrated with reference to an automotive engineering project.

Cite this paper

@article{james1998tracking,
  title = {{Tracking the right clues with exploratory data analysis}},
  author = {James, John V.},
  journal = {IEEE Spectrum},
  volume = {35},
  number = {7},
  pages = {58--65},
  year = {1998},
  doi = {10.1109/6.694359}
}

Made from the index record. Check names, capitalisation and page numbers against the paper. Need another style? AI for Research has over 10,000.

Related papers