Re: Obtaining gradients for the computation of prediction intervals

From:Leonid Gibiansky Subject:Re: [NMusers] Obtaining gradients for the computation of prediction intervals Date:Tue, 17 Sep 2002 13:10:16 -0400 Michael, Unfortunately, I do not have the paper that you cited, and cannot find it on the web. But I am not sure that you can estimate SD(observed) from anything except observations. If you have identical dosing and sampling history, SD(observed) at each time point can be directly computed from the set of observations. The formula that you mentioned is based on a particular model and looks more like a SD(predicted). Concentration at any time point can be expressed as a function of thetas and etas + error term (using previously developed population model ). Then the formula can be used to compute SD (partial derivatives should be computed at etas=0). Leonid