CBE20258 Lecture 14: NASA lecture 14
22 Mar 2016
School
Department
Course
Professor
Document Summary
N data points from a periodically replicated set. Only works if you have a really large number of data points. Take a chunk of the data, recalculate, determine how data changes. Add random noise to the data, see how things change. Particularly suitable for a small number of data points. However, if you have too few data points, dropping one could cause some chaos, making this a bad fit. If doing linear regression, you are better off doing the exact formulas, but that is more. What are the statistics of (cid:1858)((cid:1876) )? (cid:1876) =1(cid:1840)(cid:1755)(cid:1755)(cid:3533)(cid:1876)(cid:3036)(cid:3015) (cid:3036)(cid:2880)(cid:2869) (cid:1858)(cid:3037)=(cid:1858)(cid:4684)1(cid:1840)(cid:1755)(cid:1755)(cid:3533)(cid:1876)(cid:3036)(cid:3015) (cid:3036)(cid:2999)(cid:3037) (cid:4685) (cid:1858)((cid:1876) ) 1(cid:1840)(cid:1755)(cid:1755)(cid:3533)(cid:1858)(cid:3037)(cid:3015) (cid:3037)(cid:2880)(cid:2869) (cid:3364)(cid:3364)(cid:3364)(cid:3364)(cid:3364)(cid:3364)(cid:3364)=1(cid:1840)(cid:1755)(cid:1755)(cid:3533)(cid:1858)((cid:1876)(cid:3036)) (cid:3015) (cid:1858)((cid:1876)(cid:3036)) (cid:3036)(cid:2880)(cid:2869) (cid:2026)(cid:3033)((cid:3051) )(cid:2870) =(cid:1840) 1(cid:1840)(cid:1755)(cid:1755)(cid:1755)(cid:1755)(cid:1755)(cid:4684)(cid:3533)(cid:3435)(cid:1858)(cid:3037) (cid:1858) (cid:3037)(cid:3439)(cid:2870) (cid:3041) (cid:3037)(cid:2880)(cid:2869) (cid:4685) For a nonlinear estimate, this is better than: If doing linear regression, you are better off doing the exact formulas, but that is more complex. Draw n measurements at random from periodically replicated set. Do this p times (where p is a huge number)
