Standard longitudinal models (e.g., mixed models, growth curve models, etc.) are often specified with a regression term for time, and thus yield a single point estimate for the time effect. In contrast, TVEM estimates a function that represents the regression coefficient between the predictor and outcome across continuous time. Thus, unlike in more traditional models, time effects are not forced to adhere to a parametric function, but rather are flexibly estimated as nonparametric. This regression coefficient function, along with the corresponding 95% confidence interval, is best presented graphically. TVEM is particularly well-suited to investigate dynamic associations across developmental age, as demonstrated in recent studies by Evans-Polce et al. (2015) and Vasilenko and Lanza (2014). Additional technical details on TVEM, including differences between TVEM and other longitudinal models, are available elsewhere (Lanza et al., 2014; Tan et al., 2012; Vasilenko et al., 2014).
