Summary
In the process of characterizing / summarizing variables, exploratory data analysis, measures of 3 fundamental aspects come into play: location, dispersion and association.
Location essentially shows where most of the data lies, and dispersion shows whether the data making up that majority is close together. When dispersion increases, measures of location do not really say much (even if they seem to); for this very reason, both must always be present. Precisely because dispersion affects location, there is a set of possible measures that are more robust to a situation of high dispersion or to the particular dispersion of some data points. The most widely used is the median, which, unlike the average (mean), holds firm against extreme values.
Dispersion can also be misleading, because it depends on the units in which it is measured. For this reason, the coefficient of variation is recommended: it expresses the standard deviation relative to the value of the mean. As its name indicates, it is a coefficient and has no units. A value below 0.3 (30%) indicates low dispersion and, therefore, the measures of location will be reliable.
Even so, the next chapter will show that, even in this last situation, relying only on measures of location is not 100% safe. Visualization must always be used to understand the behavior of a variable or dataset, together with a frequency table.
Beyond analyzing a single variable on its own, there may also be a need to compare it with another, or even with itself under other circumstances (or at another time). For that there are measures of association, which are common in all kinds of reports. This chapter does not cover the correlation coefficient, which deserves a fuller treatment of its use and implications, given later.
Further reading
Measures of location: mean, median, quantiles and mode
- Medidas resumen (in Spanish). Kelmansky, Estadística para todos, pp. 120–125.
- Los valores que caracterizan al conjunto de datos. Aguirre, Niño and Simonetti, Estadística aplicada en las ciencias sociales y humanas (in Spanish), pp. 59–79.
- Descripción de datos. Lind, Marchal and Wathen, Estadística aplicada a los negocios y la economía (in Spanish), 15th ed., pp. 57–75.
- Medidas numéricas descriptivas. Canavos, Probabilidad y estadística. Aplicaciones y métodos (in Spanish), pp. 11–14.
- Medidas del centro de los datos (in Spanish). Introducción a la estadística, OpenStax, section 2.5.
Measures of dispersion: standard deviation, coefficient of variation and range
- Medidas resumen (in Spanish). Kelmansky, Estadística para todos, pp. 125–135.
- Análisis de la variación y asimetría. Aguirre, Niño and Simonetti, Estadística aplicada en las ciencias sociales y humanas (in Spanish), pp. 85–97.
- Descripción de datos. Lind, Marchal and Wathen, Estadística aplicada a los negocios y la economía (in Spanish), 15th ed., pp. 75–84.
- Medidas numéricas descriptivas. Canavos, Probabilidad y estadística. Aplicaciones y métodos (in Spanish), pp. 15–22.
- Medidas de la dispersión de los datos (in Spanish). Introducción a la estadística, OpenStax, section 2.7.
Measures of association: ratios, rates and proportions
- Razón, tasas y porcentajes (in Spanish). Kelmansky, Estadística para todos, pp. 15–19.