Curve fitting with 5 data points?
in [Matlab]
|
From: aaaa Vadimov on 2 Aug 2010 10:41 Hi, I have five data points and I need to show that they can be fitted by parabolic fitting. 1. Is it ridiculous to run a fit with 5 points only? 2. Is there any possible bootstrap approach (similar to linear regression with small amount of data), which can demonstrate reliability of my fit? 3. Is the curve fit the only way to show that my points are distributed in such a shape? Thanks for help, Vadim
From: Matt J on 2 Aug 2010 11:23 "aaaa Vadimov" <aaa(a)hotmail.com> wrote in message <i36le4$cv2$1(a)fred.mathworks.com>... > Hi, > > I have five data points and I need to show that they can be fitted by parabolic fitting. > 1. Is it ridiculous to run a fit with 5 points only? Depends on how much statistical variability there is in the data. Try polyfit(x,y,2) and see how good the fit is.
From: aaaa Vadimov on 2 Aug 2010 11:32 thanks I tried it and I calculated RMSE of the fit (fit points relative to original). Is there any rule of thumb what is considered good? "Matt J " <mattjacREMOVE(a)THISieee.spam> wrote in message <i36nsp$ot9$1(a)fred.mathworks.com>... > "aaaa Vadimov" <aaa(a)hotmail.com> wrote in message <i36le4$cv2$1(a)fred.mathworks.com>... > > Hi, > > > > I have five data points and I need to show that they can be fitted by parabolic fitting. > > 1. Is it ridiculous to run a fit with 5 points only? > > Depends on how much statistical variability there is in the data. Try polyfit(x,y,2) and see how good the fit is.
From: Matt J on 2 Aug 2010 11:39 "aaaa Vadimov" <aaa(a)hotmail.com> wrote in message <i36oe8$36u$1(a)fred.mathworks.com>... > thanks > I tried it and I calculated RMSE of the fit (fit points relative to original). Is there any rule of thumb what is considered good? ====== Nope. It depends on how the fit will be used, and what error magnitudes the task is tolerant to.
|
Pages: 1 Prev: Stateflow dashed transitions Next: Recording audio input using MATLAB on Mac OS X |