RE: Obtaining gradients for the computation of prediction intervals

From:Alice I Nichols [mailto:nichols@bellatlantic.net] Subject:RE: [NMusers] Obtaining gradients for the computation of prediction intervals Date:Friday, September 20, 2002 11:24 AM Mike, As a followup regarding where to find these derivatives. Here's a way I have used to reliably identify partials related to this equation. If you look at the fortran code generated by NMTRAN for your model (fsubs.for) you can find the exact variable names for both partials w.r.t. etas and partials w.r.t. epsilons. There are comment lines identifying these equations. Since there are usually several equation lines throughout the code for each partial, I suggest you follow each equation thru all the code to correctly identify these variables. After identifing these parameters you can revise your control stream to add these parameters to your output file. As Marc Gastonguay mentioned however most people will do simulations and calculate SD based on this. Alice [Alice I Nichols, PhD Hawthorne Research and Consulting, INC 132 Hawthorn Rd King of Prussia, PA 19406 PH: 610-878-9112 FX: 610-878-9113]