Aggression Data That Do Not Add Up
A 10-year-old named Marcus receives home-based services. Aggression is defined as 'hitting, kicking, or biting another person.' Frequency data are recorded by his mother on a worksheet. Last month she reported a mean of 14 episodes per day; this month she reports 3. The BCaBA notes that the mother has recently returned to work and Marcus's grandmother now provides care three days a week. Grandmother does not record data. The behavior plan calls for a dense differential reinforcement schedule and a time-out procedure for aggression. The BCaBA is preparing a quarterly progress report and is tempted to report a 79% reduction.
1. Should the BCaBA report a 79% reduction? Identify the threats to internal validity and data integrity.
Read the worked answer
No. Reporting a 79% reduction would misrepresent the data. Several threats are present. Measurement change: three of seven days are now covered by a caregiver who does not collect data, so the denominator has changed and the reported mean reflects only part of the week. Observer drift or reactivity: the mother's reporting may have changed with her schedule, and there is no interobserver agreement check. Instrumentation: a single untrained observer using an unscored worksheet is a weak measurement system. History: the return to work and change in caregiver is a major environmental change that could account for the difference independently of the plan. Regression to the mean is also possible given the unusually high baseline. The defensible report states the observed change, names these limitations, and does not attribute the change to the intervention. The BCaBA should also check whether the definition is applied consistently - 'biting' and 'hitting' may be coded differently by different caregivers.
2. What measurement and supervisory steps should precede the next quarterly report?
Read the worked answer
Steps: (1) Train the grandmother on the operational definition and the recording method, and provide a written definition with examples and non-examples. (2) Collect interobserver agreement on at least 20-30% of sessions, ideally across both caregivers, using a second observer. (3) Simplify the data sheet so recording is feasible during caregiving - partial-interval recording may be more practical than frequency if episodes are hard to count. (4) Add a social validity and contextual measure such as a daily note of caregiver and setting, so changes in the environment are visible in the data. (5) Re-baseline or conduct a brief reversal or multiple-baseline logic if the BCaBA needs to attribute change to the plan. (6) Document the measurement change in the report and note the date it occurred, so the trend line is interpretable. Until these are in place, the report should describe the data as preliminary and avoid causal claims.
3. The time-out procedure is in the plan. What should the BCaBA verify before continuing it, given the data problems?
Read the worked answer
The BCaBA must verify that time-out is being implemented as written, that it is actually a time-out from positive reinforcement in this context (i.e., the environment Marcus leaves is reinforcing), and that the procedure is not functioning as escape from demands. With inconsistent data and a new caregiver, implementation fidelity is unknown. Specific checks: observe at least two sessions with each caregiver, use a fidelity checklist, confirm the duration and release criteria, and confirm that aggression is not followed by removal from a preferred activity in a way that reinforces it. The BCaBA should also confirm that the dense DRA schedule is in place, because time-out without a rich reinforcement schedule for alternative behavior tends to be ineffective and can occasion further aggression. If fidelity cannot be established, the plan should be revised to something both caregivers can implement reliably rather than continued on paper.