Hey, Do We Really Have to Take Data?:

TLDR: This paper used a counterbalanced ABA withdrawal design to look at several aspects of data collection on instructional programs that have been taught for more than 1 month, and found that teaching assistants could accurately teach such programs without using a data sheet to guide them through the task analysis.

Abstract: This study used a counterbalanced ABA withdrawal design to look at several aspects of data collection on instructional programs that have been taught for more than 1 month. The results indicated that (a) teaching assistants could accurately teach such programs without using a data sheet to guide them through the task analysis; (b) accuracy of judging how well a student performed during a session was surprisingly similar in the data and no data conditions, though there were slight increases in accuracy under the data condition on four of the five programs; (c) all three of the teaching assistants made more accurate statements about how well a student's current performance compared with his/her previous performance when data were collected than when data were not collected, and (d) the teaching assistants also agreed more often with the instructional decisions of experts when data were collected.