How is possible to calculate error for a fitted graph?

2 次查看（过去 30 天）

显示更早的评论

Moj 2020-2-28

0
链接

此问题的直接链接

https://ww2.mathworks.cn/matlabcentral/answers/508107-how-is-possible-to-calculate-error-for-a-fitted-graph

评论： dpb 2020-2-28

Could you please guide me is there any command to calculate error for a fitted scattered data?

4 个评论
显示 2更早的评论隐藏 2更早的评论

Moj 2020-2-28

在 MATLAB Online 中打开

x = [72.01 59.9903 51.1286 45.532 36.1536 31.6631 30.16 29.0116 25.62 23.468];
sintheta = [-0.817 -0.6765 -0.8314 -0.7347 -0.6674 -0.436 0.4537 0.758 0.967 0.9972];
mdl = fittype('-1+2*sqrt(alpha./x).*exp(-beta*(x-alpha).^2)','indep','x');
fittedmdl = fit(x',sintheta',mdl,'startpoint',[23,0.06])
plot(fittedmdl)
hold on
plot(x,sintheta,'o')

dpb 2020-2-28

See the links to postprocessing fit results...it is, unfortunately, the weak link in MATLAB curve fitting.

The Curve Fitting Tool, cftool has the goodness of fit table shown if you build the model inside it which you can then export. There are a few supplied functions that take the fit object created above and return some statistics that are linked to from the postprocessing link or through the cfit object link of the output variable from fit, but none of those functions will build a nice ANOVA table or the like for you similar to SAS or other stat packages, unfortunately.

Otherwise, you're forced to build those additional results from basic principles. The PostProcessing link shows how to do some of the most basic.

As noted, big weakness in MATLAB Statistics and Fitting toolboxes in general. Pretty easy to compute the answers, not so easy for most to analyze/use the results as all you get for the most part is the most basic info.

请先登录，再进行评论。

请先登录，再回答此问题。