Model Dimension

Figure 1. Model dimension for autoregressive covariance model
Model dimension for autoregressive covariance model

The fixed effects model consists of the PROMO factor and an intercept. The random effect has identity structure and is independent across the markets identified by MARKETID.

In addition to the random effects model, WEEK has been specified as a repeated effect, with a First-Order Autoregressive covariance structure. This covariance structure is applied to the levels of LOCID*MARKETID, independently for each of the 133 store locations in the study. Since store locations are nested within markets, this is equivalent to simply specifying LOCID as the subject variable for repeated effects. However, because MARKETID is the subject variable for the random effect, it must also be specified as a repeated effects subject variable.

Next