An alternative approach to relapse analysis: Using Monte Carlo methods and proportional rates of response

Abstract

Relapse is the recovery of a previously suppressed response. Animal models have been useful in examining the mechanisms underlying relapse (e.g., reinstatement, renewal, reacquisition, resurgence). However, there are several challenges to analyzing relapse data using traditional approaches. For example, null hypothesis significance testing is commonly used to determine whether relapse has occurred. However, this method requires several a priori assumptions about the data, as well as a large sample size for between‐subjects comparisons or repeated testing for within‐subjects comparisons. Monte Carlo methods may represent an improved analytic technique, because these methods require no prior assumptions, permit smaller sample sizes, and can be tailored to account for all of the data from an experiment instead of some limited set. In the present study, we conducted reanalyses of three studies of relapse (Berry, Sweeney, & Odum, ; Galizio et al., ; Odum & Shahan, ) using Monte Carlo techniques to determine if relapse occurred and if there were differences in rate of response based on relevant independent variables (such as group membership or schedule of reinforcement). These reanalyses supported the previous findings. Finally, we provide general recommendations for using Monte Carlo methods in studies of relapse.

A Monte Carlo method for comparing generalized estimating equations to conventional statistical techniques for discounting data

“…Monte Carlo methods can be used to answer a wide variety of quantitative questions. For example Friedel, Galizio, Berry, Sweeney, and Odum (under review), used a Monte Carlo method to establish whether organisms across multiple experiments reliably demonstrated relapse behaviors or whether the organisms happen to engage in higher rates of behavior during relapse testing. In an unrelated study, Friedel, DeHart, and Odum (2017) used Monte Carlo methods to determine whether relatively small but consistent decreases in the rate of response during probe sessions were reliably caused the response dependent 100 dB tones during those sessions.…”

Further evidence of renewal in automatically maintained behavior

“…To minimize data attrition and normalize our response measure across a range of heterogeneous topographies of behaviors and recording methods (i.e., frequency, duration, percentage), we examined the magnitude of renewal using the log proportion rate of response suggested by Friedel et al (2019): where Y is the log proportion response rate, B is the response rate during session n (e.g., second postchange session), and c is a correction factor (here, the value of this factor is set to 1; see Friedel et al, 2019) to serve as a constant value to account for sessions in which no challenging behavior was emitted. This measure normalizes response rate on a session‐by session basis and facilitates the interpretation of proportional changes in response rate across individuals.…”

“…where Y is the log proportion response rate, B is the response rate during session n (e.g., second postchange session), and c is a correction factor (here, the value of this factor is set to 1; see Friedel et al, 2019) to serve as a constant value to account for sessions in which no challenging behavior was emitted. This measure normalizes response rate on a session-by session basis and facilitates the interpretation of proportional changes in response rate across individuals.…”

Social interaction can function as a reinforcer for dogs: Effects of stimulus duration and session parameters

“…To determine whether any differences in the latency to respond were meaningful, we conducted a Monte Carlo analysis, which is a type of analysis that relies on simulations of data rather than making statistical assumptions about the data. For single‐subject designs, they are useful for determining whether a specific set of behavior is meaningfully different from the baseline behavior while accounting for the variability across all of the measured behavior (Friedel et al, 2022; Friedel et al, 2019). This is achieved by simulating a new set of behavior by randomly selecting samples, with replacement, hundreds or thousands of times from all of the recorded behavior.…”

“…If the experimentally obtained sample is lower or higher than the vast majority of the random samples, it is reasonable to assume that the behavior of interest is lower or higher than the baseline. For in‐depth descriptions and examples see Friedel et al (2022), Friedel et al (2017), and Friedel et al (2019).…”