How does REGSTATS compute mse?

12 次查看（过去 30 天）
Jay Ghosh
Jay Ghosh 2014-5-8
编辑： dpb 2014-5-8
I have a very simple and seemingly silly question, but I can't figure it out. I am trying to understand why MATLAB returns a mean square error (mse) value of Inf for a perfect regression fit. Shouldn't MSE for a perfect fit be zero?
For example if I have x = [1, 3, 5] and say y = exp(x) = [2.7183 20.0855 148.4312].
Now, just for the sake of it if I do a regression using regstat as: stats = regstat(y,x,'quadratic'), it gives a perfect fit to the function (as expected, since there are 3 points) and stats.yhat are the same values as y.
stats.rsquare = 1 , which makes sense... but stats.mse = Inf. Why so ? According to the traditional formula shouldn't mse be equal to zero, since it is a perfect fit?
Any clarifications will be appreciated. Thanks!

请先登录，再回答此问题。

回答（1 个）

dpb
dpb 2014-5-8
编辑：dpb 2014-5-8
DOF = 0 so the denominator is zero--> inf for the mse.
Try with
x = [1:2:N];
y = x.^2;
instead, for example, for N=[5 7] to illustrate.

类别

在 Help Center 和 File Exchange 中查找有关 Model Building and Assessment 的更多信息

标签

Community Treasure Hunt

Find the treasures in MATLAB Central and discover how the community can help you!

Start Hunting!

Translated by