January 24, 2019
One of the challenges facing investigators who study substance abuse is determining the proper frequency with which to assess participants’ behavior. Assess too frequently and participants might quit because the burden is too high. Assess too infrequently and the data collected may be less accurate. In a forthcoming article in Addictive Behaviors, graduate student Wanjun Liu and his mentor, Methodology Center Investigator Runze Li, lead a team of researchers who examine the differences between retrospective timeline follow-back (TLFB) data and prospective daily process data. Both types of data are collected extensively in drug abuse research, and the authors propose a new method for validating TFLB against daily process data.
The authors applied their new method to both simulated and real-world data. They found that TLFB data were more or less accurate depending on the substance use being measured. Their results indicate that daily assessments are especially useful for measuring behaviors like alcohol use, which tend to vary substantially on a within-person basis. Weekly assessments, however, might be accurate for measuring behaviors like marijuana use, which tend to be more stable. Their new method was able to detect greater differences between TLFB and prospective daily process data than had been revealed in previous analyses.
For a pre-print copy of the article, please email mcHelpDesk@psu.edu.
Liu, W., Li, R., Zimmerman, M. A., Walton, M. A., Cunningham, R. M., Buu, A. (In press). Statistical methods for evaluating the correlation between timeline follow-back data and daily process data with applications to research on alcohol and marijuana use. Addictive Behaviors.