In clinical trials, we often treat site as fixed, and examine graphically the treatment effects across sites and also test for the interaction of site-by-treatment by including the interaction term in the statistical model. However, another option is to treat site as random since it is more natural. Here we test for site-by-treatment interaction using variance component. If there is a significant interaction, then the treatment effect can be estimated from the mixed effect model with both site and site-by-treatment interaction included as random effects. Participants will share their thoughts on this important issue.