tag:blogger.com,1999:blog-1909863961516003719.post4823030254717912231..comments2024-01-01T05:50:02.704-08:00Comments on joepy: Numerical Accuracy in Linear Least Squares and RescalingUnknownnoreply@blogger.comBlogger2125tag:blogger.com,1999:blog-1909863961516003719.post-23558479405987029822012-08-05T20:26:49.481-07:002012-08-05T20:26:49.481-07:00@jstults. Using orthogonal polynomials would just...@jstults. Using orthogonal polynomials would just mean a change of basis. A change of basis doesn't change eigenvalues...so the problem would still be ill-conditioned--hence small changes in the data/noise lead to huge changes in the coefficients.Anonymoushttps://www.blogger.com/profile/15953773715083084723noreply@blogger.comtag:blogger.com,1999:blog-1909863961516003719.post-55445924169887244422012-03-05T04:29:30.539-08:002012-03-05T04:29:30.539-08:00I had some fun with Filip too.
I agree: even bett...I had some <a href="http://www.variousconsequences.com/2011/08/fun-with-filip.html" rel="nofollow">fun with Filip</a> too.<br /><br />I agree: <i>even better, use an orthogonal polynomial basis instead of power polynomials.</i>Joshua Stultshttps://www.blogger.com/profile/03506970399027046387noreply@blogger.com