How fitlm computes confidence bounds?

Question

Christos Tsallis 2023-2-5

0
链接

此问题的直接链接

https://ww2.mathworks.cn/matlabcentral/answers/1906850-how-fitlm-computes-confidence-bounds

编辑： the cyclist 2023-2-7

I have used fitmid=fitlm(x1,y1), fitmax=fitlm(x2,y2), fitmin=fitlm(x3,y3) in order to return a linear regression model for my data. Then, I plot them using "plot" function. I saw that confidence bounds were also included but I would like to ask, how are they computed? Is it 95% confident interval?

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。

请先登录，再回答此问题。

Answer 1

the cyclist 2023-2-5

0
链接

此回答的直接链接

https://ww2.mathworks.cn/matlabcentral/answers/1906850-how-fitlm-computes-confidence-bounds#answer_1164165

Yes, the coefCI property of the model object output from fitlm gives the 95th percentile confidence intervals of the fitted coefficients.

2 个评论
显示无隐藏无

Christos Tsallis 2023-2-7

编辑：Christos Tsallis 2023-2-7

May you please explain what 95% confidence bound means? @the cyclist

the cyclist 2023-2-7

编辑：the cyclist 2023-2-7

fitlm estimates a coefficient for each term in the fitting equation. Call one of those coefficients beta. There is a true value of beta (that is unknown), and you have an estimate of it.

Now suppose you performed the whole experiment thousands of times, and calculated thousands of confidence intervals. You would expect that the true value beta to be inside the 95th confidence interval in 95% of those experiments.

Note that this is not quite the same as saying there is a 95% chance that the true value beta lies inside this confidence interval. But that is the nature of frequentist statistics.

This concept is not easy to understand. The wikipedia page on confidence intervals is helpful

请先登录，再进行评论。