Inference and Stable Transformations
Practical experience with fitting nonlinear models shows that for some data the model cannot be fitted at all, whereas for other data more than one solution may be found. Stable parameters enable us to describe in broad terms the characteristics of data sets likely to cause problems. For numerical analysts the answer may lie in seeking algorithms for global optimization, but for statisticians a deeper understanding of the problem is needed.
Unable to display preview. Download preview PDF.