Researchers often collect intensive longitudinal data to study the dynamic relationships between several variables across time, and often design these studies based on pragmatic choices of study feasibility. Nevertheless, researchers in the behavioral sciences rarely know the exact time period in which the dynamic relationships occur between these variables. The most common practice it to utilize data from the prior occasion to predict the next measurement occasion (i.e. lag 1). However, this decision is often done without much thought, and disregards all the other potential timescales in which these dynamic relationships might take place. Thus, the field has largely only been fixated on answering if constructs predict one another, but disregarded when constructs predict one another. The current talk will present a novel tool, the Differential-Time Varying Effect Model (DTVEM), which allows one to estimate the time periods in which variables optimally predict one another in intensive longitudinal data. The results of simulation studies show that this tool has high power and low type I error rates and can model these lags with high precision. An empirical study will demonstrate how this tool can be used to flexibly examine the dynamic relationships between several variables simultaneously across a range of potential times. Thus, this tool can not only model if constructs predict one another but can also model when these constructs predict one another across time.