Search
Program Calendar
Browse By Day
Browse By Time
Browse By Person
Browse By Room
Browse By Unit
Browse By Session Type
Search Tips
What to do in Chicago
Personal Schedule
Sign In
X (Twitter)
The analysis of single-case experimental designs (SCED) data is challenging and the debate regarding the most appropriate technique is still ongoing. Alongside this discussion, researchers working in the field are still developing and testing a variety of procedures such as multilevel models, generalized least squares regression analysis, and randomization tests. Moreover, the effect of an intervention can also be summarized as the degree of overlap between the measurements belonging to the different conditions or as difference in the mean levels of these conditions.
In the present poster we focus on the unstandardized indices, which express the difference between phases in the original metric of the dependent variable. We discuss their advantages and drawbacks in relation to standardized mean difference measures such as the d-statistic specifically designed for SCED data and non-overlap indices. We argue for the close relation to visual analysis and the ease of use and interpretation as being the main reasons for using unstandardized indices, whereas being grounded on sound statistical theory and making possible the comparisons across studies represent the main gains of using standardized indices.
The first part of the present work addresses the fact that computing and reporting effect sizes is recommended in all studies according to the guidelines of the evidence-based practices movement and answering the need to avoid publication bias. Nonetheless, we consider that the assessment of functional relations is inherent to SCED and deciding on intervention effectiveness should take into account both the magnitude of effect and the demonstration that this effect is actually due to the intervention. In order to help applied researchers assess intervention effectiveness, it is possible to complement the guidelines for systematic visual analysis outlined in 2010 by the What Works Clearinghouse panel of experts.
In the second part, we propose that this complement be based on estimating baseline trend (or level, in case trend is not present) via the well-known split-middle method. In the next step, this trend is projected it into the subsequent phase for comparing how well the actual data fit the prediction. We consider that when projecting, the variability of the baseline measurements needs to be taken into account. This variability can be operatively defined as the interquartile range, to be represented around the projected trend. That is to say, the actually obtained measurements in the intervention phase are compared to the upper and lower limits of the trend expected in case the intervention was ineffective. We comment on the advantages and limitations of this procedure for helping researchers assess functional relations.
Finally, we establish the relationship between the proposal for assessing functional relations and the extant unstandardized indices for quantifying magnitude of effect. We discuss this twofold approach with the aim to enable researchers make their own decision regarding its utility.