[EMAIL PROTECTED] (Clint Cummins) wrote: >Obviously, you can also estimate the above model with a linear >regression package, by collecting terms on the parameters to be >estimated
Thanks, but at present I don't really want a package. I want to understand the process, which is why I chose the simplest possible constrained situation, namely a linear fit on 2 intervals with just a single continuity constraint. What's bugging me is that I thought I understood the principles, but with test data I can't make it work. I want to use a model in which the sum of squared deviations (ssd) over the whole data set is minimised (splines are piecewise minimisations, I thought?). Anyway, I set up the partial deriv equations for minimising the sums of squares as shown in my previous post. The results are not right. Is there an obvious error in the way I set up the simultaneous equations? The books and web sites I've found so far are too generalised for me to follow - I need to understand the simplest model first. ross . . ================================================================= Instructions for joining and leaving this list, remarks about the problem of INAPPROPRIATE MESSAGES, and archives are available at: . http://jse.stat.ncsu.edu/ . =================================================================
