Recently there has been an increase in the collection of intensive longitudinal data (ILD), or data collected at many occasions over a relatively short period of time. Instead of investigating slow-developing characteristics as in more traditional longitudinal studies, ILD allow researchers to look into daily patterns, variation, or quickly-changing subject attributes. The increased availability of remote data collection devices such as smartphones and activity monitors has led to a corresponding increase in the collection of such data. When participation in a certain activity is of interest, applied researchers are often interested in using ILD to model the time until individuals participate in such activities, leading to survival and recurrent event analses. While the activity of interest can be time-stamped exactly, predictors are often recorded regularly and on different scales than the outcome activity, and can include left- and interval-censoring. This paper presents an application to an ILD data set of times of marijuana use, with predictors such as craving and anxiety, and addresses issues with the irregularity of the data collection of marijuana use and predictors.