Measurement and Data Interpretation
Selecting a measure
The first decision is what to measure and how. Continuous measurement captures every instance: frequency/count, rate (count per unit time), duration (total or per occurrence), latency (time to onset), and inter-response time. Discontinuous methods sample behavior within intervals - partial-interval (any occurrence), whole-interval (behavior throughout), and momentary time sampling (behavior at the instant the interval ends). Discontinuous methods systematically distort estimates: partial-interval tends to overestimate, whole-interval tends to underestimate, and momentary time sampling is usually the least biased for duration-like behaviors.
Agreement and accuracy
Interobserver agreement (IOA) quantifies consistency between observers. Total count IOA = smaller count ÷ larger count × 100. Interval-by-interval agreement = agreements ÷ (agreements + disagreements) × 100. Trial-by-trial and exact-count methods are stricter. A frequent trap: high IOA does not guarantee accuracy - two observers can agree and both be wrong if the definition is poor.
Graphing and visual analysis
Line graphs are standard: time on the x-axis, behavior on the y-axis. Visual analysis examines level, trend, and variability within and across phases. Celeration describes change in rate over time. Watch for ceilings/floors, artifacts from discontinuous recording, and confounded phase changes. Report trend direction and stability before claiming an effect.
Worked example
Observer A counts 20 responses, Observer B counts 25. Total count IOA = 20/25 × 100 = 80%. Rule of thumb: 80% is a common minimum, but 90%+ is preferred for high-stakes decisions, and IOA must be paired with a clear, observable definition.