Toward a reporting standard for codebooks and data dictionaries describing complex data

Google map

Abstract: 

Data sharing is increasingly important for reproducibility and replicability of scientific findings (Tenopirl et al., 2011). One of the main challenges of utilizing shared data is that it may be difficult for researchers not privy to the data collection process to interpret a given data set (Arslan, 2019). Ecological momentary assessment (EMA) approaches exacerbate this problem. The unique ability of EMA approaches to deliver adaptive assessment and intervention poses unique challenges for sharing and reporting data. For example, EMA studies often utilize adaptive assessment via skip logic or triggered surveys, resulting in complex patterns of missing data; often use intricate and sometimes data-reliant longitudinal sampling schemes with uneven timing; and make a number of small choices about the layout, size, anchoring, and number of items that may or may not influence results. Codebooks and data dictionaries provide a primary avenue to make data usable to outside researchers by providing both metadata (information about the data itself; Buchanan et al, 2021) and paradata (information about the process of data collection; Groves & Heeringa, 2006). However, few guidelines or standards exist to help researchers communicate this kind of information effectively, especially for complex studies using ambulatory data.

In this talk, we propose a set of principles for data dictionaries and codebooks for studies including EMA data and adaptive study designs. We will discuss requirements needed to ensure minimum usability; reporting of missing data from complex study designs; and propose standardized definitions for key data sharing terminology. Recognizing the risk that these guidelines create significant burden for the researcher, we also introduce a proof-of-concept automated tool to quickly and easily generate EMA codebooks and data dictionaries that are both machine-processable and human-readable.