How can I get a matrix for such a for loop ? I want to create a matrix with each of the i values, because now I only get always the last value for i.

I've two matrixes G and Rm and I want to regress one column of the G matrix on one column in the Rm matrix. When I made this in the down way, I see in command window the different values, but I only get as variable the last i output.

for i=1:6 
fitlm(G(:,i),Rm(:,1))
fitlm(G(:,i),Rm(:,2));
fitlm(G(:,i),Rm(:,3));
fitlm(G(:,i),Rm(:,4));
fitlm(G(:,i),Rm(:,5));
fitlm(G(:,i),Rm(:,6));
fitlm(G(:,i),Rm(:,7));
end

Is there the possibility to solve this in another way function or so, to get in the end a matrix with the different variables ? If possible can I create a matrix with the regrossor coefficients ?

Thanks for help.

Answer 1

[jgg]

You have to save the output at each step. The ans variable is being overwritten. For example

 C = struct();
 for i=1:6 
  C.(strcat('R1C',num2str(i))) = fitlm(G(:,i),Rm(:,1))
  C.(strcat('R2C',num2str(i))) = fitlm(G(:,i),Rm(:,2));
 ... (etc)
end

Would store your linear models in the elements of the structure label RiCj corresponding to the element you're using. Is this what you wanted to do?

(I notice you could also loop over the Rm variable instead of hard-coding that as well).

[Fox]

Hi, thanks. For the help. I wanted to have a matrix C of the Coefficients at the end. You know therefore a solution ? If I want to loop Rm , do I need an additional loop in the old one ?

[jgg]

In order to get the coefficients, you have to extract them from the linearmodel object, which is not what your original code is doing. The fitlm function does not return a coefficient vector; it returns a linearmodel object.

However, this object contains all the relevant information you'd want. Click on it in the variable viewer to see the different statistics.

Let's assume your linear model has the same number of coefficients k. Then, you could do this as follows:

 C = zeros(k,6*7);
 count = 1;

 for i=1:6 
   for j = 1:7
    lm = fitlm(G(:,i),Rm(:,j));
    coeffs = lm.Coefficients(:,1);
    C(:,count) = table2array(coeffs);
    count = count + 1;
 end
end

if your models have different coefficient numbers, you'll have to adjust this a little bit.