75
Another limitation of this study relates to the fact that data from the clinic-referred and non-referred groups were collected differently. Observational data for the clinicreferred group were collected on two separate visits, one week apart. In an attempt to reduce attrition, observational data for the comparison group were collected on the same day, with a half-hour interval between each observation. The data for the comparison group may be less stable in that they were collected on only one day, and in that the dyads may have become more tired over two consecutive 25-minute observations. Determining the best method to collect observational data (e.g., length of session, number of sessions) is an important empirical question that requires investigation.
This research has several implications for the use of direct observations as an
assessment tool. In future projects utilizing DPICS II, the researchers may choose to limit the number of categories coded to a subset of the entire system that is relevant to their research question. DPICS II was designed for this purpose (Eyberg et al., 1994). By limiting the number of categories, the coding will be less demanding on the observers for both their attention and the time required to observe. Limiting the categories also is likely to increase the reliability estimates (Suen & Ary, 1989).
In addition, the complexity of the coding system, particulary its exhausive coding of each consecutive behavior, lends itself to a lag sequential analysis. Other researchers employing this technique have found patterns of behavior that differentiate types of clinicreferred dyads. Dumas (1984), for example, evaluated conditional probabilities for parent and child aversive behavior and found that mothers who were unsuccessful in parent training were more aversive and indiscriminate in their behavior toward children. It is