loop for OLS using fitlm

Question

Alexis Villacis 2018-12-18

0
链接

此问题的直接链接

https://ww2.mathworks.cn/matlabcentral/answers/436240-loop-for-ols-using-fitlm

回答： Felipe Valenzuela Soto 2019-4-11

I am trying to estimate standar errors of my betas using bootstraping. I am using fitlm.

My Y and x are two vectors of dimensions 202x1. How can I program the loop to save the betas in a matrix? I should have 2 betas for each of the 1000 models.

My code is as follows:

%Get Initial values

OLS=fitlm(x,Y)

%save residuals

res=OLS.Residuals.Raw;

%generate y hat

ypred=predict(OLS1,x)

%bootstrap 1000 draws of the residuals with replacement

resboot=res(randi(202,202,1000))

%generate 1000 new draws of Y

yboot=ypred+resboot

% Do the fits

for i = 1:1000

mdl= fitlm(x,yboot(:,i));

end

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

Felipe Valenzuela Soto 2019-4-11

0
链接

此回答的直接链接

https://ww2.mathworks.cn/matlabcentral/answers/436240-loop-for-ols-using-fitlm#answer_370198

Hi Alexis,

Maybe you already solve it, but i think you should first create two zeros vector of 202x1 to asign the estimated betas and residuals from the loop using the fitlm function. I have used this method in my codes for yield curve term structure estimation, more related to financial application. I hope it works for you.

beta = zeros(size(X,1),1);

residuals = zeros(size(X,1),numel(X));

%loop

for i = 1:size(X,1)

EstOLS = fitlm(X, Y','Intercept", false); %Or true in you want

beta (i ; :) = EstOLS.Coefficients.Estimate';

residuals(i ; :) = EstOLS.Residuals.Raw';

end

Good luck!

Felipe Valenzuela.