Editorial

Abstract

Event History Data

Vis. Methods

Vis. Tools
  Case Study 1

Comparing Pencils

Further development

Difficulties

New tools

Acknowledgements

References

Case Studies Index

3. The development of visualisation tools

We begin by considering the types of variable which might exist in an event history dataset. First, there are variables as proxies for time, such as age of the individual, and calendar date. Second, there will be usually a large number of time-varying variables, representing changes of state or value over time. These may be continuous (such as earnings/week), ordinal (such as educational level reached) or categorical (such as type of crime committed). Third, there are time-constant variables, which do not change over time, such as the gender or ethnic group of the individual. Lastly, there are pure events, either internal to the history of the individual (such as gaining a driving licence) or common to all individuals in the study (such as a change in government).

How do we best represent an individual history? When examining the sleeping and feeding patterns of a recently born child, Cleveland (1994) considers a simple block diagram. However, Cleveland's idea cannot easily be extended to examine multiple histories. Barry, Walby and Francis (1989) devised a circular 'tulip' diagram, consisting of a number of concentric rings, each representing a different variable, with colour, shading and width of the rings being used to represent changes in values over time. However, such a diagram places undue visual emphasis on the variables contributing to the outermost rings, and this is not a desirable characteristic. Examples of each of these displays can be found on our web site (http://www.cas.lancs.ac.uk/alcd/visual/)

We consider instead the metaphor of a pencil, with each ot its faces representing a different variable, and its length representing the length of the event history. Calendar time runs along the length of the pencil. Changes in each of the face variables can be represented by size, value, texture, colour, orientation or shape. Categorical variables can be represented by a set of colours, textures or patterns.

3.1 Case Study 1 - Employment in Kirkcaldy

As an example, Figure 1 shows a perspective view of a typical pencil representing an employment event history for a married couple taken from a retrospective survey of 188 living in Kirkcaldy, Scotland in 1985 (for further details of the dataset see Francis and Fuller, 1996). The event history for the couple begins at the date of marriage at the left hand side of the diagram, and continues until the survey date.

Figure 1: A perspective view of a pencil representation of the life history of a married couple. Time runs from left to right, starting at date of marriage and finishing at the survey date. Each face of the pencil represents a different variable. The top face represents the employment history of the wife, the middle face that of the husband, and the bottom face the age of the youngest child in the household. Explanation of the colours can be found in the text.

The three faces, taken in order, and proceeding clockwise from the top of the pencil, are female employment status, male employment status, and the age of the youngest child in the household. Employment status for both husband and wife is coded 0 (dark blue) for not working and 1(mid-blue) for working. The age of the youngest child in the household is coded 3 (green) for no children, 4 (yellow) for a child aged under 1, 5 (red) from age 1 to under 5, 6 (magenta) from age 5 to under 11 and 7 (light green) from age 11 to under 16. In this history, we can see that the husband never worked throughout the survey period. However, the wife worked up until the birth of her first child, then stopped work until her youngest child was aged 10, when she returned to work, becoming unemployed for a while in the 1981 recession.

Other analysts might want to examine other variables over time, such as migration, housing tenure, educational level of the husband and wife, and so on. The faces can be reassigned to new variables or extra faces can be added to the pencil display.