Re: WT as significant covariate on peripheral volume

Hi Ayyappa, As Sven mentioned, it would make the most sense for it to be a significant covariate on both central and peripheral volumes. Just because the objective function value drops more doesn't mean that you shouldn't include the effect of weight on central volume in your final model. The peripheral volume yielding a greater change on OBJ could mean several different things including: • Physiologically, the portion of the body that is not at rapid equilibrium with plasma is more dependent on weight than the portion of the body that is at rapid equilibrium. • Study design(s), you have more data at times after the in the peripheral compartment is dominant (i.e. after the alpha phase of elimination). • Mathematically, the error model has a smaller relative variance at the times with the peripheral compartment is dominant. All three of these are likely inter-mixed to give the true answer. As a general note, just because something isn't statistically significant doesn't mean that you can't include it in your model. Sometimes, you will want to include biologically logical components in your model independent of the statistics. Many people will include weight on all volumes linearly and weight^0.75 on clearance without testing the significance because of the biological relevance. Thanks, Bill