All event-based IOA algorithms have in common the analysis of concordance on frequency counts and event records. These measures consist of (a) total number, b) partial concordance in intervals, c) exact match, and (d) DOA trial by trial algorithms. After a brief overview of each event-based algorithm, Table 1 summarizes the strengths of the four event-based algorithms for behavioral reliability analysis considerations. Suppose a research team collects frequency data for a target response of 15 observations of 1 m (see Figure 1). Interval for the IOA interval. In short, the Interval by Interval method evaluates the proportion of intervals in which the two observers agreed to determine whether the targeted response occurred. Note that this includes agreement on the existence and absence of a target response. This is calculated by adding the total number of agreed intervals by the sum of the number of agreed and non-agreed intervals. As might be expected, this approach often leads to high convergence statistics. As Cooper et al.

(2007) report, this is particularly true when partial interval readings are used. In the data examples in Figure 2, observers disagree on the first and seventh intervals, resulting in an interval-by-interval compliance value of 71.4% (5/7). For continuously collected data, the calculation of percentage overreality was developed using time window analysis. Intervals of one second are imposed on the data streams of two observers and comparisons of seconds are made between them. If both data sets show an event (for discrete behaviors) or a second of common occurrence (for behaviors measured over time), this is counted as a concordance. Every second that a single record contains an event or behavior is a disagreement. The percentage of match is calculated by diverging the number of chords by the number of chords plus disagreements. MacLean et al.

(1985) understood that their algorithm was too strict for discrete event data. Therefore, they recommended allowing tolerance for counting agreements by broadening the definition of a chord to observations when one observer records an event within the ± t seconds of the other observer. . . .