Forms of Longitudinal Data

• The simple term `longitudinal' is ambiguous:
  • Pure panel longitudinality:
    • State at (approximately) evenly-spaced discrete observation points
  • Continuous observation - discretely or continuously changing state
  • Discrete:
    • All dates of change of state recorded
    • Within the panel period or retrospectively
  • Continuous: `Calendar' model
    • Totals, summaries or samples on a frequent (e.g., monthly, quarter-hourly) basis
    • e.g., monthly income data, time-use diary data
  • the first can be regarded as `repeated measurement', the second as `event history', the third `state history'

• The three forms have advantages and disadvantages
• The two continuous observation forms are clearly richer: pure panel is intermittent, has gaps (cf Blossfeld and Rohwer)
• However, panel is very expensive, but very broad, detailed
• In continuous observation models, the type of variable is important: something that changes only occasionally and stays constant for considerable periods is well captured in an event history (e.g., place of residence, number of children, job)
• Something that changes continually (e.g., temperature) or can sensibly be aggregated over short periods (e.g., income, rainfall) is suited to state history