Bootstrap linear regression MSE

7 次查看（过去 30 天）

Moh Aljoh 2019-7-27

0
链接

此问题的直接链接

https://ww2.mathworks.cn/matlabcentral/answers/473720-bootstrap-linear-regression-mse

编辑： Adam Danz 2019-7-28

I am trying to use bootstrap on the MSE, R-squared output generated from linear regression model. However, I am having trouble figuring out how to set it up with the correct arguments.

I tried to do something like:

X1 = outcomes;
X2 = modcomes;
mdl = LinearModel.fit(X1, X2);
 resid = outcomes - modcomes;
% Simple bootstrap example
N_Boot = 1000;
SSE = zeros(N_Boot,1);
R_Sqrd = zeros(N_Boot,1);
for i = 1:N_Boot
     [foo_b , GoF_b] = LinearModel.fit(modcomes, outcomes + resid);
     SSE(i) = GoF_b.sse;
     R_Sqrd(i) = GoF_b.rsquare;
end
mean(SSE)
std(SSE)
mean(R_Sqrd)
std(R_Sqrd)

1 个评论
显示 -1更早的评论隐藏 -1更早的评论

Adam Danz 2019-7-28

编辑：Adam Danz 2019-7-28

If you're using matlab r2013b or later, you should use fitlm() instead of LinearModel.fit(). They have virtually the same inputs and both produce the LinearModel object. The model contains a field "Residuals" that contains (you guessed it) the residuals of the model. There is no documented second output and I haven't tried doing that myself so I'm not sure what's in the 2nd output in your code.

What are modcomes and outcomes?

请先登录，再进行评论。

请先登录，再回答此问题。