Using Variance Components to Assess a Random Effect

In previous analyses, a grocery store chain studied the relationship between customer shopping behavior and the amount spent. There is, however, a lot of store-to-store variation that reduces your ability to estimate the effects of these behaviors. By adding the store location as a random effect, you can reduce the amount of unexplained variation, thus increasing the accuracy of your estimates of other model terms.

This information is collected in grocery_1month.sav. See the topic Sample Files for more information. Use the Variance Components procedure to fit a model with fixed and random effects to the amounts spent, and estimate the contribution of the store to the variation.

Next