Getting to know your variables: The foundation for a good working relationship with your data

• Activity Number: 168
• Type: Topic Contributed
• Date/Time: Monday, August 5, 2013 : 10:30 AM to 12:20 PM
• Sponsor: Section on Statistical Education
• Author(s): Jane Miller*+
• Companies: Rutgers University
• Keywords: Communication ; Statistical analysis ; Study design ; Variables

Abstract

Interpretation of measures of distribution and association or multivariate coefficients depends on the meanings, units, and context of the variables involved. Study design and methods of data collection affect key attributes of the sample and variables. Neglecting to become acquainted with one's variables can lead to incorrect results and the need to redo analysis. In this paper, I describe an exercise to help researchers get to know their variables before analyzing them, making use of documentation on the study from which the data are drawn, the data set itself, and published literature on the topic under study. I start by discussing the importance of identifying the context, unit of analysis, and limits on the analytic sample to suit one's research question. I then describe and illustrate a series of steps to be conducted for each variable in the analysis, including becoming familiar with the conceptually valid range of values, levels of measurement, units and categories, missing values, and observed distribution of values. I explain how these steps are an integral part of conducting and writing about an application of a statistical analysis to a real-world question.