Implementing Automated Nonparametric Statistical Analysis on Functional Analysis Data: A Guide for Practitioners and Researchers

Abstract

Functional analysis (FA) is an integral component of behavioral assessment and treatment given that clinicians design behavioral treatments based on FA results. Unfortunately, the interrater reliability of FA data interpretation by visual analysis can be inconsistent, potentially leading to ineffective treatment implementation. Hall et al. (2020) recently developed automated nonparametric statistical analysis (ANSA) to facilitate the interpretation of FA data and Kranak et al. (2021) subsequently extended and validated ANSA by applying it to unpublished clinical data. The results of both Hall et al. and Kranak et al. support ANSA as an emerging statistical supplement for interpreting FA data. In the present article, we show how ANSA can be applied to interpret FA data collected in clinical settings in multielement and pairwise designs. We provide a detailed overview of the calculations involved, how to use ANSA in practice, and recommendations for its implementation. A free web-based application is available at https://ansa.shinyapps.io/ansa/.

Supplementary information: The online version contains supplementary material available at 10.1007/s40614-021-00290-2.

Keywords: Data interpretation; Functional analysis; Statistical analysis; Visual analysis.

Fig. 1

Case example from Kranak et al. (2021). The graph shows responses per minute plotted across sessions for each condition in a multielement design. The table beneath the graph shows the statistical output from ANSA

Fig. 2

Flowchart to illustrate the process ANSA employs to determine whether or not a condition is differentiated

Fig. 3

Case example to illustrate implementation of ANSA on FA data collected in a pairwise design. The graph shows responses per minute plotted across sessions for each test-control condition pair. The table beneath the graph shows the statistical output from ANSA

Fig. 4

Case example to illustrate implementation of if-then rules when no conditions are differentiated. The graph shows responses per minute plotted across sessions for each condition in a multielement design. The table beneath the graph shows the statistical output from ANSA. Note that an ignore condition was included to test for automatic reinforcement

Fig. 5

Case example to illustrate implementation of ANSA on FA data collected in a pairwise design. The graph shows responses per minute plotted across sessions for each test-control condition pair. The table beneath the graph shows the statistical output from ANSA. Note that a tangible (edible) condition was included to determine whether problem behavior was maintained by this specific form of tangible reinforcement