Residual variability

From nami_iida@rnd.fujisawa.co.jp Sun Jun 22 20:15:32 1997 Subject: Residual variability Dear NONMEM Users; Recently, I have a question about my interpretation of residual variability in using 1st-order method. I'll show the simplest example. Suppose all subjects under population analysis get the constant infusion, and the population model below is chosen. Css=(Rate/(TVCL+etaCL))*(1+eps) In this case, I think, this model linearized to the following model if 1st-order method is used to analyze. Css=Rate/TVCL+(-Rate/(TVCL**2))*etaCL+Rate/TVCL*eps It means that the standard error of Css caused by the residual variability is ALWAYS CONSTANT to be "Rate/TVCL*sigma" even if etaCL variates. For example, if Rate=10 and TVCL=1, and there is a subject whose etaCL is 0.1, his standard error of Css according to residual variability is regarded to be "10/1*sigma", NOT TO BE "10/1.1*sigma". Could someone show me if I make a mistake or not? Thank you. Nami Iida