Chapter 8 A Reproducible Exploratory Data Analysis

Exploratory data analysis is a reproducible investigation of what the data can support. This chapter develops a complete workflow for orientation, data-quality assessment, univariate and multivariate analysis, missing-data and outlier decisions, and the conversion of exploratory findings into documented analytical consequences.

Learning outcomes

After completing this chapter, you should be able to:

• conduct a structured data-quality assessment;
• select purposeful univariate, bivariate, and multivariate analyses;
• investigate and justify decisions about missing values and outliers;
• separate observed findings from possible explanations;
• document an EDA so that it can be reproduced; and
• convert exploratory findings into analytical questions and project decisions.

Key terms

• Exploratory data analysis (EDA): A structured investigation of data quality, distributions, relationships, patterns, limitations, and analytical possibilities.

• Data quality: The degree to which data are accurate, complete, consistent, timely, valid, and suitable for their intended use.

• Distribution: The pattern of values taken by a variable, including their frequency, centre, spread, shape, and extremes.

• Missingness: The amount, pattern, and process through which expected data values are absent.

• Outlier: An observation that is unusually distant from or inconsistent with other observations under a relevant definition.

• Unit of analysis: The entity represented by one observation in an analysis.

• Univariate analysis: The examination of one variable at a time.

• Bivariate analysis: The examination of a relationship or comparison involving two variables.

• Multivariate analysis: The examination of relationships involving three or more variables.

• Reproducibility: The ability to recreate results from documented data, code, settings, transformations, and decisions.