Effects of data sampling on graphical depictions of learning

Area of Science:

• Behavior Analysis
• Applied Behavior Analysis (ABA)
• Research Methodology

Background:

• Discrete-trial programming is a common ABA teaching strategy.
• Efficient data collection is crucial for monitoring progress in ABA.
• Continuous data collection can be time-consuming; sampling methods are often considered.

Purpose of the Study:

• To compare the accuracy of continuous versus discontinuous data-collection methods in discrete-trial programming.
• To evaluate the impact of different trial and session sampling strategies on estimating learning metrics.
• To assess the time-saving potential of data sampling methods.

Main Methods:

• Analysis of archival data sets using discrete-trial programming.
• Implementation of trial sampling (e.g., 1st 5, 1st 3, 1st trial).
• Implementation of session sampling (e.g., every 2nd, 3rd, 5th session).

Main Results:

• Trial sampling systematically underestimated sessions/days to mastery but overestimated sessions/days to the first independent response.
• Session sampling systematically overestimated both sessions/days to mastery and sessions/days to the first independent response.
• Time-savings analysis indicated minimal time savings across all sampling methods.

Conclusions:

• Both trial and session sampling methods introduce systematic biases in estimating learning progress.
• The potential time savings from data sampling are likely minimal and may not outweigh accuracy concerns.
• Continuous data collection may be more reliable for accurately tracking progress in discrete-trial programming.