Inexperienced observers, as a group, have not been far behind experienced observers, with precision and precision. A newcomer (DK) proved to be as accurate and precise as the best experienced observers. With training similar to that described for experienced observers, it is reasonable to expect beginners to all become experts. The results of the calibration make it possible to offer more targeted training (e.g.B. systematic errors that lead to undervaluation suggest that greater attention to the registration task could be encouraged). Note that we recommend above all to train observers and not to apply a posteriori correction of systematic distortions to values obtained on the basis of the calibration of individual observers. The potential benefits of calibration for applied behaviorists are hypothetical. Such speculations must be thoroughly evaluated by research which shows our field the advantages and disadvantages of a multitude of methods of obtaining the reference values essential for calibration. An important question concerns the circumstances in which it might be reasonable to foresee that calibration could replace the Interobserver agreement as an accepted method for quantifying the quality of our data. Consider an applied study in which the baseline response rate is between 6 and 8 responses per minute.
Reference values for two or more meetings during the initial value could be determined from sets of criteria. The reference values would then be obtained from several additional sessions, while the response rate during treatment decreased (e.g. B between 6 and 1 response per minute) and again in case of low response in the storage phases (e.g.B. less than 1 response per minute). Up to 10 session reference values would be required to be compared to measurements made by observers to calibrate the study data. The most economical method for obtaining reference values would be the use of a single expert observer to measure reference values for calibration (as in Sanson-Fisher et al., 1980; Wolfe et al., 1986). Observer data would be represented graphically as the background data of the study (as is currently the case) as well as calibration statistics, either numerically or graphically, instead of compliance measures known to Interobserver. The question of whether researchers, experts and behavioral editors would accept such a method is empirical. An inter-observer agreement has been reported to replace accuracy. Interobserver compliance is calculated by comparing two continuous recordings recorded simultaneously by independent observers.
The Interobserver agreement can be seen as a poor substitute for accuracy, as we cannot determine to what extent one of the observer`s recordings is a “true” representation of interest. Nevertheless, the use of inter-observer agreement calculation methods is considered essential to ensure the specificity of behavioural definitions as refined during the initial development of an observation system, to ensure that observers react homogeneously to defined behavioural responses and to assess the impact of observer training. The information presented so far makes it possible to predict how observers (with 95% certainty) would obtain results when measuring pinch behavior with rates that are within the limits of calibration samples (i.e. 0 to 8 responses per minute). . . .