Curve fitting with more than 3 variables

Question

Tania 2014-6-4

0
链接

此问题的直接链接

https://ww2.mathworks.cn/matlabcentral/answers/132421-curve-fitting-with-more-than-3-variables

评论： John D'Errico 2018-12-2

Hi! I am doing curve fitting with matlab at the moment and would like to do it with more than 3 variables.Is that possible and how can i do it? please see the screenshot, I can not add any further variables there. Thank you!

1 个评论
显示 -1更早的评论隐藏 -1更早的评论

Mahdi 2014-6-4

Do you mean you want to do it in a higher dimension than 3D? Or more polynomial coefficients?

请先登录，再进行评论。

请先登录，再回答此问题。

Answer 1

dpb 2014-6-4

2
链接

此回答的直接链接

https://ww2.mathworks.cn/matlabcentral/answers/132421-curve-fitting-with-more-than-3-variables#answer_139458

That's correct; the cftool GUI is limited to curves or surfaces.

You can define other more general models you wish with with either GeneralizedLinearModels tools in the Statistics Toolbox (under 'regression') or can always write the normal equations directly and use the backslash ("\") operator to solve.

Details on what your inputs are and the desired model could lead to more specific responses.

0 个评论
显示 -2更早的评论隐藏 -2更早的评论

请先登录，再进行评论。

Answer 2

dpb 2014-6-5

2
链接

此回答的直接链接

https://ww2.mathworks.cn/matlabcentral/answers/132421-curve-fitting-with-more-than-3-variables#answer_139537

编辑：dpb 2014-6-5

在 MATLAB Online 中打开

Don't know about web sites; the Matlab doc starts under

doc

Open the Statistics Toolbox section, select "Regression and Classification" and then "Parametric Regression". Start at the bottom with the Concepts section and work your way thru to the Linear Models which is probably all you need.

ADDENDUM:

There's a little more to it than the following, but the general idea for indicator variables is easily seen in the following

https://www.moresteam.com/whitepapers/download/dummy-variables.pdf

For setting up contrasts between more locations, there are some restrictions on the dummy encoding. Any good text on regression will address it.

3 个评论
显示 1更早的评论隐藏 1更早的评论

dpb 2014-6-13

"More", I don't know would be a good way to put it; it's an alternative choice, yes. I really have no direct experience in the specific area of regression trees, sorry.

Don't know you've anything to lose by trying both and seeing which does a better job for your purposes on the data you have. You'll want to avoid making any generic conclusions from just a sample dataset results, obviously...

Tania 2014-6-13