Help generating bootstrap standard errors for parameters generated from fminunc?

3 次查看（过去 30 天）

jlr 2019-2-5

0
链接

此问题的直接链接

https://ww2.mathworks.cn/matlabcentral/answers/443390-help-generating-bootstrap-standard-errors-for-parameters-generated-from-fminunc

编辑： jlr 2019-2-5

I am trying to generate standard errors using a bootstrap for parameters generated from fminunc. The issue is that the ojective function in fminunc must be scalar, but I have B different objective functions for each repetition of the bootstrap--in this case fun2b has the following dimentions (1,1,B). I think if I could pull each of the b elements from fun2bs, I could estimate this fine, but I don't see how to pull elements from a parameterized function.

Apologies if this is a stupid question, as I'm sure is clear from the code below, I am new to matlab.

My code is as follows:

for b=1:B

y(:,:,b)=datasample(wavemovs,100); **** create B copies of bootstrapped data dependent variable***

x(:,:,b)=datasample(x7ss,100); **** create B copies of bootstrapped data independent variables***

fun2bs=@(bs)probitboot(bs,y,x) **** function that generates the objective function which is **dependent on parameter vector bs and returns a 1*1*B matrix

fminunc(fun2bs,beta0,opts)

end