“If you torture the data enough, it will confess to anything,” purportedly said Ronald Harry Coase, winner of the Nobel Prize in Economics. As behavior analysts, we do our best to let the data “show what it shows,” and then respond. But what if our method of visual display isn’t totally honest with us?
The claim: social stories reduce behavior
In this webinar (above), Dr. Rick Kubina and Dr. Amanda Kelly looked at data from an intervention for Peter, a 7-year-old boy. The data came from “Social Story Intervention: Improving Communication Skills in a Child with an Autism Spectrum Disorder,” published in A Focus on Autism and Other Developmental Disabilities.Social stories are well-known interventions, widely said to reduce challenging behavior in children with autism spectrum disorder. Similarly, the article authors suggested that their intervention DID reduce behaviors: “As in previous studies, the present study showed a decrease in the participant’s frustration behaviors (i.e., crying, falling, hitting, and screaming) when the social story was introduced” Adams, Gouvousis, VanLue, & Waldron, 2004, p. 92).
The question: did they really?
But then, Rick recharted the data on a Standard Celeration Chart. He used two features available on the Standard Celeration Chart… and discovered that the social stories may not have been as effective for Peter as claimed.
Feature 1: Level and Level Change
Level displays the average rate of responding. The average can be found several ways (we like the geometric mean). Rick compared the level change between the baseline, intervention, removing intervention, and reinstating intervention for each of the 4 undesirable behaviors.
With level, you can look at each condition… and look within and between each condition. Level quickly shows how much the rate changes. A clinically significant change is 2 or greater. But in this intervention, the level change wasn’t visually impressive. And quantification showed the intervention was not clinically significant: level varied 1.66 or less between conditions.
Feature 2: Celeration (rate of change)
Looking at trend, you can see if the data is flat, increasing, or decreasing. But celeration gives us quantified numbers that tell exactly how fast behavior changes. When Rick analyzed at the social story data with celeration, a new picture emerged.
He again compared celeration (increase and decrease) between conditions. During the social stories intervention, challenging behaviors accelerated. In fact, acceleration from baseline to social stories for “episodes of crying” was x1.83 (which is close to the clinically-significant minimum of 2). If the speed of something doubles, that’s significant — especially when the goal is behavior reduction. Similar data emerged with the other undesirable behaviors.
The answer: not for Peter!
This social stories intervention did not meaningfully reduce the level of (average) episodes of behavior. While social stories appeared to work initially with some behaviors, Peter later responded positively to the withdrawal of social stories. Further intervention effects show Peter’s behavior worsened upon reintroduction of the social stories.
The Standard Celeration Chart revealed critical information about the intervention. Our field says that if you have data, you should be compelled to change to follow the data. But if the data is distorted by a non-standard graph, it’s difficult to make master-level data-driven decisions. That’s why Standard Celeration Chart shouts out the statistical value of change — so you can help clients by following the data.
Already watched the webinar video? Take BACB Type 2 CEU quiz here. Worth 1 credit.
*Paid content by Chartlytics.