dc.description.abstract | Obtaining contextually stable estimates of child challenging behavior is of interest to researchers who design interventions to decrease child challenging behavior. However, there is little empirical guidance on measurement systems needed to obtain sufficiently stable estimates of challenging behavior, specifically the minimal length of observation sessions, minimal number of raters, and behavior sampling method. We conducted a generalizability study in which we partitioned error variance between four facets of our system: session length (5, 10, 15 mins), behavior sampling method (partial interval and timed event), number of raters (1, 2), and number of occasions (1, 2, 3, 4 observations). With information gained from the generalizability study, we conducted several decision studies to identify optimal design characteristics. The number of occasions produced the greatest variance in our scores. We conducted 6 decision studies for each combination of length of observation and behavior sampling method to examine the length of sessions, number of raters, and number of occasions needed to be averaged in order to obtain acceptable stability estimates (i.e., g coefficient of 0.70). Timed event sampling with 15 five-min occasions averaged across 3 raters was the least resource-intense design that produced acceptable levels of stability. Research is needed to increase the stability and validity of measuring child challenging behavior in unstructured contexts. | |