Knowledge Base

Problem

A reviewer or user needs to understand and explain how Observer XT calculates inter-rater reliability for point events, specifically when a tolerance window (e.g., 1.5 seconds) is specified.

Details

Affects Observer XT reliability analysis using the Frequency/Sequence method with a user-defined tolerance window. Point events have no duration, so their agreement logic differs from state events.

Solution

For **point events** using the **Frequency/Sequence** reliability method with a tolerance window:
1. Observer XT compares events of the **same behavior code** across the two observations.
2. A point event is counted as an **agreement** when its recorded onset time falls within the specified tolerance window (e.g., ±1.5 seconds) of the corresponding event in the other observation.
3. Events of the same code that **cannot be matched** within the tolerance window are counted as **disagreements**.
4. The reliability score is calculated from the ratio of matched (agreed) events to total events across both observations.
5. When reporting this to reviewers, specify: the method used (Frequency/Sequence), the tolerance window value, and that unmatched events outside the window were treated as disagreements.