On the definition of latent state-trait models with autoregressive effects: insights from LST-R theory

In longitudinal studies with short time lags, classical models of latent state-trait (LST) theory that assume no carry-over effects between neighboring occasions of measurement are often inappropriate, and have to be extended by including autoregressive effects. The way in which autoregressive effects should be defined in LST models is still an open question. In a recently published revision of LST theory (LST-R theory), Steyer, Mayer, Geiser, and Cole (2015) stated that the trait-state-occasion (TSO) model (Cole, Martin, & Steiger, 2005), one of the most widely applied LST models with autoregressive effects, is not an LST-R model, implying that proponents of LST-R theory might recommend not to apply the TSO model. In the present article, we show that a version of the TSO model can be defined on the basis of LST-R theory and that some of its restrictions can be reasonably relaxed. Our model is based on the idea that situational effects can change time-specific dispositions, and it makes full use of the basic idea of LST-R theory that dispositions to react to situational influences are dynamic and malleable. The latent variables of the model have a clear meaning that is explained in detail.