How to assess and take into account trend in single-case experimental design data

ABSTRACT

One of the data features that are expected to be assessed when analyzing single-case experimental designs (SCED) data is trend. The current text deals with four different questions that applied researchers can ask themselves when assessing trend and especially when dealing with improving baseline trend: (a) What options exist for assessing the presence of trend?; (b) Once assessed, what criterion can be followed for deciding whether it is necessary to control for baseline trend?; (c) What strategy can be followed for controlling for baseline trend?; and (d) How to proceed in case there is baseline trend only in some A-B comparisons? Several options are reviewed for each of these questions in the context of real data, and tentative recommendations are provided. A new user-friendly website is developed to implement the options for fitting a trend line and a criterion for selecting a specific technique for that purpose. Trend-related and more general data analytical recommendations are provided for applied researchers.

Trial registration: ClinicalTrials.gov identifier: NCT04560777

KEYWORDS:

• Single-case experimental design
• trend
• regression
• split-middle

Disclosure statement

No potential conflict of interest was reported by the author(s).

Notes

1 For instance, the What Works Clearinghouse (2022) makes a recommendation about how to quantitatively analyse the data, but both this recommendation and the disregard for visual analysis have been questioned (Kratochwill et al., 2021; Maggin et al., 2022).

2 See the later section entitled “Comparison between the Quantifications and the Clinician’s Judgment” for further discussion of the term “deterioration” and is broader meaning, beyond the sign of the quantification.

3 It is common in the context of analytical approaches applicable to SCED data, such as multilevel modeling (Moeyaert, Ferron et al., 2014) and Bayesian analysis (Natesan Batley, 2022), and it possible when carrying out experience sampling (Weermeijer et al., 2022).

4 The number of data points to remove, in the context of such an exploratory study, could being fixed a priori or determined by visual analysis as in response-guided experimentation.